MCP Server

Clustr provides an MCP (Model Context Protocol) server that lets AI assistants like Claude Desktop, Claude Code, and Cursor interact with your Clustr network directly through natural language. Ask your AI assistant to search opportunities, manage target accounts, or import contacts without leaving your workflow.

Setup

Configure Claude Code

Clustr connects two ways. Both register a server named clustr — pick one.

Sign in (OAuth)
API key

Recommended. No API key, no Node.js. Run this in your terminal:

claude mcp add --transport http clustr https://api.clustr-ai.com/mcp

Then run /mcp in Claude Code and choose Authenticate to sign in with Clustr in your browser. The tools appear once you approve access.

For headless or shared setups. Requires Node.js 18+ (download). Run this in your terminal:

claude mcp add clustr -e CLUSTR_API_KEY=your-api-key -- npx -y @clustr-ai/mcp-server

Replace your-api-key with a key from your API settings page. The server is available immediately, no sign-in and no restart needed. npx downloads the server automatically on first run.

Other clients (Claude Desktop, ChatGPT, Cursor)

Add Clustr as a custom connector using the HTTP endpoint, then sign in with Clustr:

https://api.clustr-ai.com/mcp

In Claude or ChatGPT, open Settings → Connectors → Add custom connector, paste the URL, continue, and approve access. The same endpoint accepts a static API key via an Authorization: Bearer your-api-key header for clients that support custom headers.

Available Tools

Once connected, your AI assistant has access to the following tools:

New tools need a reconnect, not a reinstall or a re-login. OAuth and custom-connector clients talk to https://api.clustr-ai.com/mcp directly, and the API-key install (npx -y @clustr-ai/mcp-server) is a stdio bridge to that same endpoint — it declares no tools of its own and fetches the list from the server. So there is never an npm version to bump and never a token to refresh.What a client does cache is the list it fetched when it connected. Restart it, or run /mcp in Claude Code and reconnect Clustr, and the new tools appear. Your credential is untouched either way.

Tool	Description
`list_account_research`	New in v1.11.0. List the account research reports that already exist in your workspace, newest first, whether they were started in the web app, the API or MCP. Free
`get_account_research`	Changed in v1.30.0. When the last run on the company failed, answers `status: FAILED` with `failure_reason` and `reason_code` instead of “no research found”. Read an existing report: the warm buyers at the company, the intro paths into each (which connector, through which shared company, for how long, and why the path is relevant), the best first move and the magic link. Look it up by `report_id`, `company_linkedin_url` or `company_name`. Never starts a run and uses no credit
`start_account_research`	Changed in v1.30.0. A refusal says why it was not started, and a run that fails later is reported by `get_job_status` and on the magic link. Start a new account research run on a target company and get a shareable magic link immediately. Uses one credit. If research already exists for the company, it returns that report with `already_exists: true` and starts nothing; pass `refresh: true` to run it again. Renamed in v1.9.0 (was `start_abm_search`, which still works but is no longer listed)
`get_account_research_link`	Fixed in v1.11.0. Retrieve the magic link for an existing report without running it again. Now also finds research started in the web app. Renamed in v1.9.0 (was `get_abm_search_link`, same)
`import_connectors`	Renamed in v1.2.0 (was `import_contacts`, which no longer resolves). Import LinkedIn connectors for referral opportunity discovery. The call starts the import and returns its `job_id`; connectors appear within a couple of minutes and their warm intros once the import finishes, usually within 10 to 15 minutes. New in v1.32.0: `mark_companies_as_clients` adds the connectors’ current companies to your client list once the import finishes.
`search_opportunities`	Changed in v1.20.0. Takes `evidence`, the Warm intros views: `recommended` (an interaction or a recommendation, or strength Good and above) or `all` (also the weak paths where the two only worked together). Filter on `has_interaction`. Changed in v1.67.0: `hide_big_company_overlaps` and `big_company_max_employees` were removed. Paths whose prospect is also one of your connectors are now included. Without `exclude_client_companies`, results follow your workspace’s client companies setting. Search opportunities by connector, company, or across your entire network. Filter on `cities` (the prospect’s city, full names such as San Francisco) and on `icp_fit` (strong, partial, none). Filter on `tag` with tag ids: tags from the same group match any of them, tags from different groups must all match. Owner filters take owner ids from `list_owners`, and an owner email still resolves. Company filters match any role a person currently holds, so a search for a fund finds the partner filed under one of its portfolio companies. Add `include_all_positions` to see every job each person holds or held. Changed in v1.45.7: the description says what `total` counts: paths, the People list’s count with no filters (the same as `get_network_stats`’ `total_opportunities`). Changed in v1.52.0: `account_tag` filters on the tags of the prospect’s account
`search_companies`	Fixed in v1.35.0. Search your own companies, the ones your network has warm intros into, by name (`q`) or `domain`. It does not search LinkedIn; `search_linkedin_companies` does
`list_connectors`	List people in your network with opportunity counts. Changed in v1.45.7: `total` counts the people in your network with at least one path, the same as `get_network_stats`’ `connectors`. Changed in v1.52.0: `account_tag` (the connector’s account) and `path_tag` (at least one of their warm intros)
`set_connector_owner`	Changed in v1.15.0. Set who owns a connector, and therefore everything it produces, with no reimport. The same thing `owner_email` sets at import time. `push_to_crm` also writes the owner to the connector’s CRM contact; the result’s `crmPushSummary` says in one sentence what happened there
`get_owner_field_mappings`	New. Which field of a connected CRM holds the account owner (`company`), a connector’s owner (`contact`) and the CS owner (`company_cs`), and which side may change each. See Get Owner Field Mappings
`set_owner_field_mappings`	New. Map those fields, checked against the CRM’s live schema. The CS owner has no default field and syncs like the account owner once mapped; Defense competitor alerts mention it. Admin only
`get_tag_group_crm_mappings`	New in v1.54.0. Which property of a connected CRM each tag field syncs with, and which side owns it
`set_tag_group_crm_mappings`	New in v1.54.0. Set those mappings; the list replaces the CRM’s whole set. Admin only
`unassign_connector_owner`	New in v1.5.0. Clear a connector’s owner. Opportunities already assigned keep theirs
`delete_connector`	New in v1.6.0. Remove a connector and delete every opportunity that came through them, for people who have left. Irreversible, and flagged destructive so clients confirm first
`get_network_stats`	Get an overview of your network: connectors, opportunities, top companies. Changed in v1.45.7: counts what the lists show: `total_opportunities` is `search_opportunities`’ total with no filters and `connectors` is `list_connectors`’ total
`get_job_status`	Changed in v1.30.0. A `FAILED` job carries `failure_reason` (why, in plain English, with what to do) and `reason_code` (a stable code). New: `overlaps_pending` is `true` while a finished job’s warm-intro overlaps are still being worked out in the background, so its opportunity count can still grow. Changed in v1.55.0: a `COMPLETED` import whose derivation did not finish carries `derive_incomplete`, `derive_incomplete_reason` and `derive_note`. Get the status and progress of an import or account research job. A `QUEUED` job starts automatically once imports already in progress free up capacity
`list_failed_import_items`	New in v1.27.0. List the people an import could not bring in, each with its LinkedIn URL, a plain-English reason (“Profile not found or removed”, “LinkedIn URL is not valid”, “Temporary error, retry”), a stable `reasonCode`, the raw error and whether a retry re-processes it. See List Failed Import Items
`list_jobs`	Changed in v1.30.0. List import and account research jobs with optional status filter (`QUEUED`, `PROCESSING`, `COMPLETED`, `FAILED`, `CANCELLED`). A `FAILED` job carries `failure_reason` and `reason_code`. Each job carries `overlaps_pending` (`true` while a finished job’s overlaps are still draining). Changed in v1.55.0: `derive_incomplete`, `derive_incomplete_reason` and `derive_note` on a `COMPLETED` import whose derivation did not finish. Changed in v1.45.7: `page` and `page_size` are honoured
`get_opportunity`	New in v1.10.0. Read one opportunity by id, including the notes and tracking set on it. The way to read back what `update_opportunity` wrote
`cancel_job`	New in v1.18.0. Stop a queued or running import. What it has not started is skipped; what it already imported stays
`get_matching_settings`	New in v1.19.0. Read what your workspace counts as a path: the evidence kinds, what a small company is, how old a shared job may be, what same place means, and whether client companies are hidden. See Matching settings
`update_matching_settings`	New in v1.19.0. Change your matching settings. Workspace admins only; only the fields you send change, and every change is logged
`update_opportunity`	Changed in v1.15.0. Update one warm intro: its own owner (`owner_id` from `list_owners`, or `unassign_owner` to clear) and its notes. That is the owner the People table shows; `set_connector_owner` sets the connector’s owner instead and moves every opportunity the connector has. Intro tags are tags, set with `tag_records`
`list_tag_groups`	Changed in v1.65.0. Your three tag fields (Account tags, Connector tags, Intro tags) with their tags. A tag marked `closes` closes the warm intro it is on
`tag_records`	Add, remove or replace tags on up to 100 accounts, people or warm intros at once
`list_tags`	Every tag, flat, each with its `group_id` and `subject_type`. `list_tag_groups` shows them under their fields
`create_tag`	Changed in v1.65.0, breaking. Create a tag in a record type’s field. Takes `subject_type` (`account`, `person` or `path`), `name`, `color` and `closes`
`update_tag`	Changed in v1.15.0. Works on every tag, not only connector tags. Rename it, recolour it, or change `closes`; `position` is gone. Tag ids did not change, so a call on a connector tag works as before
`delete_tag`	Changed in v1.15.0. Deletes any tag. Every record carrying it loses it
`reorder_tags`	New in v1.54.0. Set the order of a field’s tags
`set_connector_tags`	Changed in v1.65.0. Set the Connector tags on one connector, by their LinkedIn id. Tags are connector-level, so this affects every warm intro that connector produced. Several tags are fine
`list_owners`	Changed in v1.15.0. List the owner directory, with each owner’s opportunity count and whether they are a workspace member. When a CRM is connected each entry carries `crmSeat` — provider, `hasSeat`, `deactivated` — which is whether a `push_to_crm` write can name that owner in the CRM
`add_owners`	New in v1.48.0. Add one owner or up to 2,000 at once, e.g. a rep list from your CRM, as `owners: [{email, first_name, last_name}]`. Sends no email and grants no access. Idempotent: an address that is already an owner is left alone and a deleted one comes back. Replaces `create_owner` (retired; a call says to use this) and `import_owners` (still works as another name for this)
`update_owner`	New in v1.4.0. Update an owner’s email or name
`delete_owner`	New in v1.4.0. Delete an owner. Every opportunity they hold becomes unowned, and the count is returned
`list_accounts`	Changed in v1.14.0. The accounts your network reaches, hottest first: each one’s heat (`burning`, `warm`, `cool`, `cold`), last signal and 30-day signal count, fit grade (`A+`..`D`) with every criterion behind it, tags, connector count and best path strength. Filter by `fit`, `temperature` or `tag`; sort by `heat` (default), `fit`, `path`, `recency` and more. This is your account list; `search_companies` finds a company by name or domain and says nothing about how hot it is. Changed in v1.38.0: with Competitor Monitoring, each account a watched competitor has touched carries `competitor_activity`, filterable with `competitor`, `competitor_pressure`, `competitor_last_touch`, `competitor_side`, `competitor_active_days` and `competitor_pitch_days`, and sortable with `lastCompetitorActivity`, `lastCompetitorPitch`, `competitorPitches` and `pressure_path`. Changed in v1.52.0: `path_tag` keeps accounts with at least one warm intro carrying those tags
`list_account_tags`	Your Account tags, in display order. Each tag carries the `group_id` of its field
`create_account_tag`	Changed in v1.65.0. Create an Account tag. `group_id` is gone. A name already in the field is rejected rather than duplicated
`update_account_tag`	New in v1.11.0. Rename a tag, recolour it or move it. Only provided fields change; assignments reference the id, so a rename breaks nothing
`delete_account_tag`	New in v1.11.0. Delete an account tag. This also removes it from every account carrying it, and cannot be undone
`set_account_tags`	New in v1.11.0. Replaces an account’s tags with the set you send, not an add. An empty list clears them
`create_target_accounts`	Fixed in v1.23.1. Add target accounts for opportunity prioritization. Each account takes a `company_name`, a `company_linkedin_url`, an optional `tier`: the name your workspace uses, and an optional `tag`: one of your account tags. An unknown tier or tag rejects the whole request rather than being created, so call `list_tiers` and `list_account_tags` first. A company already on the list still gets the tag. Before v1.23.1 every call was rejected. Changed in v1.45.8: `company_linkedin_url` is required, the company’s LinkedIn page (`linkedin.com/company/acme`); an account without a valid one is not created and is listed in `rejected` with the reason
`list_target_accounts`	Changed in v1.10.0. List or search your target accounts. Rows now carry `tier`, `tier_id` and `company_id`; an empty `company_id` means the account has not matched a company yet and cannot be tiered. Changed in v1.45.7: takes `page` and `page_size` (ordered by name, with `total` and `has_more`); with neither, the whole list. Changed in v1.45.9: `linkedin_lookup` is `not_found` on an account whose LinkedIn company page could not be found after two lookups; it may be a wrong or outdated URL
`set_target_account_tier`	New in v1.10.0. File one target account under a tier, by name or by `tier_id`. An empty value clears it. Fails with a conflict when the account has not matched a company yet — fix its LinkedIn URL first
`list_tiers`	New in v1.10.0. Your workspace’s account tiers in rank order (position 0 is highest). Tier names are yours, so read them before writing one
`delete_target_account`	Fixed in v1.23.1. Remove a target account by company `name`. Before v1.23.1 every call was rejected
`list_crm_account_lists`	The company lists in a connected CRM that your target account list can follow: HubSpot lists, Salesforce list views, Attio saved views. Takes `provider`
`bind_target_accounts_to_crm_list`	Follow a CRM list with your target account list, so the nightly sync keeps them in step. The list’s companies are added; nothing you entered is removed. Takes `provider` and `list_id`
`unbind_target_accounts_from_crm_list`	Stop following a CRM list. The accounts it added are removed; accounts you entered stay. Flagged destructive
`list_client_companies`	New in v1.14.0. Your client companies, one entry per company, with every source behind it (manual, csv, or a CRM) and whether it is marked not a client. Filter by `source`, search with `q`
`get_client_companies_summary`	New in v1.14.0. How many client companies you have, how many nobody in your network works at, the count per source, and each CRM’s last client sync
`add_client_companies`	New in v1.48.0. Add one client company or up to 1,000 at once, as `companies: [{name, linkedin_url, domain}]`. `linkedin_url` must be a LinkedIn company page, not a person’s profile; `domain` is optional. A company without a valid one comes back in `invalid` with its row number and the reason, and the others are added; companies already listed are counted in `alreadyListed`. Replaces `add_client_company` and `import_client_companies` (retired; a call says to use this)
`remove_client_company`	New in v1.14.0. Remove one manual or csv row from your client companies. Flagged destructive. CRM rows cannot be removed, because the next sync would bring them back: use `mark_not_client`
`mark_not_client`	New in v1.14.0. Mark the company behind a `source_row_id` not a client. Overrides every source, including later CRM syncs. Flagged destructive
`undo_not_client`	New in v1.14.0. Undo a not a client mark, by `exclusion_id`
`get_client_rule`	New in v1.38.0. Read who counts as a client in a connected CRM (`provider`): a company property condition, a won deal (HubSpot `hs_is_closed_won`, Salesforce `IsWon`, Attio the deal stages you pick), or both. See CRM rule
`list_client_rule_fields`	New in v1.38.0. The CRM company properties a client rule can use, with their options, and Attio’s deal stages
`preview_client_rule`	New in v1.38.0. Count the companies a draft client rule would select, with a few names, without saving it
`set_client_rule`	New in v1.38.0. Set who counts as a client in a CRM: a property condition, a won deal, or both, matched with `any` or `all`. Starts a pull of that CRM’s clients
`sync_client_rule`	New in v1.38.0. Pull a CRM’s clients now with the saved rule
`delete_client_rule`	New in v1.38.0. Remove a CRM’s client rule; the companies it marked leave your client list. Flagged destructive
`list_excluded_companies`	The companies excluded from your workspace, each with what it is matched on (company id, LinkedIn id, LinkedIn slug or domain)
`exclude_companies`	New in v1.48.0. Exclude one company or up to 500 everywhere: each leaves every list, search, account, warm intro, count and export and stops being enriched. Takes `companies: [{label, domain, linkedin_url, linkedin_numeric_id, company_id, reason}]`, at least one key per company. Invalid companies come back in `invalid` with the reason, companies already excluded are counted in `alreadyExcluded`, and `excludedCompanies` returns the new exclusions with their ids. The list holds at most 500 companies. Not the same as `mark_not_client`. Flagged destructive. Replaces `exclude_company` (retired; a call says to use this)
`undo_exclude_company`	Remove one exclusion, by its id from `list_excluded_companies`
`list_icps`	Changed in v1.25.0. List your Ideal Customer Profiles: industries, employee counts and countries. Revenue is no longer a criterion
`create_icp`	Changed in v1.25.0. Create an ICP from industries, employee counts and countries. No longer takes `revenue`: revenue is missing for many companies over $10M, so a revenue rule excluded good fits. A `revenue` value sent anyway is ignored
`update_icp`	Changed in v1.25.0. Update an ICP; only provided fields change. No longer takes `revenue`, and saving an ICP clears any revenue range it still held
`delete_icp`	New in v1.4.0. Delete an ICP
`list_competitors`	List monitored competitors with slot usage. Changed in v1.42.0: slots count monitored people, whatever the competitor: `used`, `total`, `paused`, `pending` and `plan`
`preview_competitors`	New in v1.42.0. Check LinkedIn company URLs before adding them: the company, whether it is already registered, the Sales and CS people found and the slots they would take. Changed in v1.47.0: takes `title_groups` and lists the first page of people found (`people`). Changed in v1.49.0: each person carries `group`, the role group of their title
`search_linkedin_companies`	New in v1.35.0. Search LinkedIn company pages by name, to pick the right company before `register_competitor`
`resolve_linkedin_company`	New in v1.38.0. Resolve a LinkedIn company URL, vanity name or numeric id to the ids `register_competitor` takes
`register_competitor`	Register a competitor to monitor. Starts its first scan. Changed in v1.35.0: takes `linkedin_numeric_id` and `linkedin_vanity_name` from `search_linkedin_companies`, and refuses a LinkedIn URL that names no company page. Changed in v1.42.0: takes several companies at once with `companies`; no competitor limit, the people found are monitored up to the free slots and the rest are Paused; an already registered company is refused with `COMPETITOR_EXISTS`. Changed in v1.45.0: takes `website_url`, the marketing site its customers are found on (`https://<domain>` by default). Changed in v1.47.0: takes `title_groups` (the roles followed) and `people` (each preview person followed or skipped)
`update_competitor`	Update a competitor’s name, domain or LinkedIn URL. Changed in v1.35.0: `status` is gone; competitors can no longer be paused. Changed in v1.45.0: takes `website_url`
`delete_competitor`	Delete a competitor and stop monitoring it
`list_competitor_team`	New in v1.35.0. The people watched at a competitor, ranked executives first, with their title, role and last activity. Changed in v1.42.0: each person has a function, a status (monitored, paused, pending, left), a source, an added-on date and 30-day interactions; `status` filters the list
`search_linkedin_people`	New in v1.60.0. Find someone on LinkedIn by name, to follow them at a competitor: an AE the job titles missed, an investor or a partner
`add_competitor_team_member`	New in v1.35.0. Watch one more person at a competitor, by LinkedIn profile URL. Changed in v1.42.0: or by `employee_id` from `list_competitor_employees`; takes a slot, `NO_SLOT` at the limit. Changed in v1.60.0: takes `relationship` (employee, investor, partner, advisor, other); the team check never marks someone added by hand as having left
`remove_competitor_team_member`	New in v1.35.0. Stop watching one person; the next team search does not add them back. Changed in v1.42.0: frees their slot
`update_competitor_team_member`	New in v1.42.0. Monitor or pause one person: resume a paused person, validate a joiner, or keep someone after a role change
`validate_competitor_joiners`	New in v1.42.0. Monitor or ignore the joiners a team refresh found, in bulk. Changed in v1.59.0: kept for compatibility; joiners are followed automatically, so it normally finds nobody waiting
`swap_competitor_people`	New in v1.42.0. Pause one monitored person and monitor another in one step
`list_competitor_people`	New in v1.42.0. Everyone tracked across all competitors, filtered by competitor, function and status, with an accent-insensitive name search
`list_competitor_employees`	New in v1.42.0. A competitor’s other employees, to pick people to monitor
`get_competitor_slots`	New in v1.42.0. Your slots: used, total, paused, pending and plan
`get_competitor_team_titles`	New in v1.42.0. The job titles the team search looks for, each with its function
`set_competitor_team_titles`	New in v1.42.0. Replace the title list; it applies from the next team search
`reset_competitor_team_titles`	New in v1.42.0. Go back to the default title list
`list_competitor_customers`	New in v1.36.0. The companies your competitors sell to, with the evidence, an estimated renewal window, and whether each is in your CRM or is your own customer. `competitor_id` narrows it to one competitor. Changed in v1.45.0: each customer is labelled Confirmed, Likely or Possible from its signals (case study, testimonial, LinkedIn post, logo wall, CS interactions), with the contacts found; only Confirmed and Likely by default (`include_possible` for the rest); filter by `label` and `signal`. Changed in v1.49.0: rows carry the CRM account owner and CS owner; `order=standing` lists your own customers first, then the companies in your CRM, and `owner` keeps one owner’s. See Their customers
`set_competitor_customer_verdict`	New in v1.36.0. Confirm or reject one of those customers; a rejected one never comes back. Changed in v1.38.0: `verdict: clear` removes a verdict
`add_competitor_customer`	New in v1.60.0. Add a company you know buys from a competitor, for example an account you won from them. It is saved as confirmed and stays listed with no evidence; `set_competitor_customer_verdict` with `verdict: clear` undoes it
`set_competitor_customer_renewal_date`	New in v1.45.0. Record when a competitor customer renews; it replaces the estimated window. `clear` removes it
`list_competitor_customer_people`	New in v1.49.0. The people at one competitor customer: `contacts` found there through the competitor, and `buyers`, up to 20 people there your network reaches whose title is one of your buyer titles, each with the best way in (connector and strength) and whether they are in your CRM. `buyers_note` says why `buyers` is empty
`lookup_competitor_customers_by_company`	New in v1.38.0. Which competitors each of a set of your accounts (`company_ids` from `list_accounts`) is a customer of. Changed in v1.45.0: Confirmed and Likely customers only, with the label
`get_competitor_customer_settings` · `update_competitor_customer_settings`	New in v1.45.0. Renewal tasks: a CRM task for the account owner when a confirmed competitor customer is one of your accounts, and when it is due. Off by default; admins change it
`get_competitor_crm_properties` · `set_competitor_crm_properties`	New in v1.60.0. Write which competitors a company is a customer of, the confidence and the last competitor engagement date onto the companies in your CRM. Each is off until an admin turns it on. See Set Competitor CRM Properties
`get_competitor_site`	New in v1.45.0. The website a competitor’s customers are found on, when it was last read (about once a month), and the pages read
`preview_competitor_website`	New in v1.45.0. Check a website before saving it as a competitor’s `website_url`: where it ends, its title, and the domain that will be read
`recrawl_competitor_site`	New in v1.45.0. Read a competitor’s website again now. Admins only, once a day per site
`list_competitor_interactions`	New in v1.4.0. The Signals feed — every competitor-rep interaction with a person at one of your accounts, with the ICP, CRM and critical verdicts. Filter by competitor, window, engagement type, rep role and seniority, and by CRM scope — including `crm=customer` for the accounts you already own. Both directions: `direction` is `outbound` (the rep engaged the person’s post) or `inbound` (the person engaged the rep’s own post), and filters on it. Changed in v1.49.0: `account` keeps one account’s interactions. Changed in v1.61.0: `intent` keeps the interactions on a post that shows buying intent (`any`, or one intent)
`list_competitor_interaction_prospects`	New in v1.4.0. The same feed folded to one row per engaged person. Takes the feed’s window, competitor, CRM, critical and type filters, not `target`, `order`, `roles` or `seniority`
`export_competitor_interactions`	New in v1.4.0. Your entire interaction history in one call
`list_competitor_accounts`	New in v1.4.0. Every account competitors are touching, ranked, with your CRM ownership and open-deal detail. Changed in v1.42.0: `scope=attack` ranks the accounts that are not your customers and `scope=defense` lists your customers, by interactions over `period`. Changed in v1.46.0: `scope=all` is one list of both plus every company that is a customer of a competitor you watch, narrowed with `yours` and `customer_of`; every scoped row carries `customer_of` and `we_own`. Changed in v1.47.0: `in_crm`, `activity_only` and `owner` filters; rows carry `people_engaged`, `latest` and the CRM owner and CS owner. Changed in v1.49.0: rows carry `people` (up to 10 of the people engaged, newest first, with their newest interaction and its pitch fields) and `pitches`; `order=standing`, `types` and `pitch` on every scope. Changed in v1.60.0: rows carry `in_market` and `in_market_facts`; `in_market=true` keeps the accounts in market. Changed in v1.61.0: rows carry `latest_intent`, the newest buying intent at the account; `intent` keeps the accounts with buying intent
`export_competitor_accounts`	New in v1.46.0. The competitor account list in one call, unpaged: `scope` (required: `attack`, `defense` or `all`), `period`, `competitor_ids` and, with `all`, `yours` and `customer_of`. See Export Competitor Accounts
`export_competitor_customers`	New in v1.63.0. Every competitor’s customers in one call, unpaged, with the filters of `list_competitor_customers`. Each row adds `people_you_can_reach`, `sales_owner` and `cs_owner`. See Export Competitor Customers
`list_competitor_account_interactions`	New in v1.4.0. One account’s full competitor history
`get_competitor_account`	New in v1.42.0. One account, per competitor: interactions over 30 and 90 days, the salespeople involved, the last interaction and a 13-week series
`list_competitor_summaries`	New in v1.4.0. Header tallies for every tracked competitor, plus `totals` (`accounts_touched`, `at_risk`) counting each account once across competitors
`get_competitor_detail`	New in v1.4.0. One competitor’s battlecard: ranked accounts, reps, activity over time
`list_competitor_runs`	New in v1.4.0. In-flight and recent scans, with status
`snooze_competitor_interaction`	New in v1.4.0. Hide one interaction until a deadline you set
`dismiss_competitor_target`	New in v1.4.0. Permanently silence one person
`undismiss_competitor_target`	New in v1.4.0. Reverse a dismissal
`list_dismissed_competitor_targets`	New in v1.4.0. The people you have dismissed
`ignore_competitor_interaction`	New in v1.42.0. Ignore one interaction for good: it never alerts and leaves every count
`unignore_competitor_interaction`	New in v1.42.0. Undo an ignore
`list_ignored_competitor_interactions`	New in v1.42.0. The interactions you ignored, for the undo
`log_competitor_interaction_to_crm`	New in v1.4.0. Write a critical signal onto the matched CRM company as a note
`get_competitor_monitoring_schedule`	New in v1.4.0. When the automatic collection runs: every two weeks, on a weekly slot (Mondays at 11:00 Europe/Paris by default) that each competitor takes every other time, once you have added competitors. Changed in v1.54.0: every two weeks
`update_competitor_monitoring_schedule`	New in v1.4.0. Set the day and time of the slot, and whether it runs at all. The cadence is pinned to weekly, the slot. Changed in v1.51.0: takes `day_of_week`
`run_competitor_monitoring`	Collect now, outside the schedule, for one competitor or all of them. Throttled per workspace
`get_competitor_delivery_config`	New in v1.4.0. Where finished reports are delivered
`update_competitor_delivery_config`	New in v1.4.0. Set delivery channels and the engagement policy. Changed in v1.38.0: `routes` sends each audience (open deals, the rest of the CRM, accounts outside it) its own part of the report
`get_competitor_alerts`	New in v1.35.0. Where competitor alerts go: per use case (Attack, Defense) the Slack channel and the email (to the owner plus extra addresses), the webhook, the Attack threshold (10 interactions in 30 days by default, per competitor or across all of them) and the history window (14 days by default). Changed in v1.45.3: Attack covers your target accounts and every other non-customer account, each alert labelled Target account or Other account; `attack_other_accounts` says whether the other accounts alert. Changed: one post, one email per recipient and one webhook call per alert, naming the owner read live from your CRM; no direct messages. `account_id` also names the owner an Attack alert on that account mentions
`update_competitor_alerts`	New in v1.35.0. Changed: set the Attack and Defense Slack channels and emails (`attack_email_enabled`, `attack_email_recipients` and the Defense pair), the webhook (URL, secret, on or off), the Attack threshold and its scope (`attack_threshold_scope`), the history window (`alert_history_days`), whether Attack also alerts on other accounts or on target accounts only (`attack_other_accounts`, new in v1.45.3), or turn alerts off. With no channel, no email and no webhook, no alert is sent. The old recipient, DM and single-channel fields are ignored. Admin only
`preview_competitor_alert`	New. See what a Defense or Attack alert would send, built from a real account with recent competitor activity, with a draft of the `update_competitor_alerts` fields laid over the saved ones: the message, the owner it names, where it goes and how many accounts would alert today. Saves and sends nothing. See Preview Competitor Alert
`test_competitor_alert_webhook`	New. Send a signed sample competitor alert to your webhook, the saved one or a URL you pass. See Webhook payloads. Admin only
`set_connector_slot`	New in v1.16.0. Keep a connector fresh: put them in a connector slot the refresh automation re-ingests
`clear_connector_slot`	New in v1.16.0. Take a connector out of their slot
`get_person_note`	New in v1.24.0. Read your workspace’s note on a person, connector or prospect, with who last edited it and when. The note the app shows on the person’s page and when someone asks for an intro through them
`set_person_note`	New in v1.24.0. Write the note on a person. It replaces the note there was, so read it first to add to it; an empty note clears it. The person must be one of your connectors or appear in one of your warm intros
`clear_person_note`	New in v1.24.0. Remove the note on a person. Flagged destructive
`get_connector_refresh_schedule`	New in v1.16.0. When the connectors in a slot are re-ingested
`update_connector_refresh_schedule`	New in v1.16.0. Set when the refresh runs, and whether it runs at all
`run_connector_refresh`	New in v1.16.0. Re-ingest every connector in a slot now
`get_failed_run_alerts`	New in v1.16.0. Whether you are emailed when a workflow fails, and who receives it
`update_failed_run_alerts`	New in v1.16.0. Turn failed-run emails on or off and set the recipients
`get_broken_integration_alerts`	New in v1.17.0. Whether you are emailed when a connection stops working, and who receives it
`update_broken_integration_alerts`	New in v1.17.0. Turn broken-integration emails on or off and set the recipients
`get_data_store_options`	New in v1.62.0. The tables each connected data store (Supabase, Neon, Snowflake, BigQuery) receives: Accounts, Connectors and competitor monitoring
`update_data_store_options`	New in v1.62.0. Turn those tables on or off on a data store (`export_accounts`, `export_connectors`, `export_competitors`). Turning one on syncs at once. Name `provider` when several are connected
`list_slack_channels`	New in v1.31.0. The channels in your connected Slack workspace that Clustr can see, with whether Clustr is in each one and what it may do there: `none`, `read`, `write` or `read_write`. See Slack channel access
`update_slack_channel_access`	New in v1.31.0. Set, per channel, whether Clustr may read it, post to it, both or neither, and the default for every other channel. Workspace admins only
`join_slack_channel`	New in v1.31.0. Add Clustr to a public Slack channel. A private channel needs an `/invite` in Slack. Workspace admins only
`list_competitor_reports`	List competitor-monitoring reports, newest first
`get_competitor_report`	Get one competitor report by ID, including full attack/defend data
`list_saved_views`	New in v1.12.0. List your saved views for one table: `entity` is people, accounts or connectors (required)
`list_workflows`	New in v1.12.0. List your Workflows: view, trigger, member count, and last run
`list_signals`	New in v1.29.0. Every LinkedIn reaction and comment on your warm intro paths, and every post by a person you track, newest first (`order: "asc"` for oldest first). Filter by `since`/`until` (a day or an ISO-8601 time), `families` (`network`, `activity`, `prospect`), `origins` (`prospect`, `network`) and `types` (`reaction`, `comment`, `post`, `repost`); narrow by the person’s account (`industries`, `employees`, `revenues`, `countries`, `tiers`, `owners`, `tags`) and seniority (`seniority_ranks`); page with `offset`. Each row carries `opportunity_id` for `get_opportunity`, and since v1.30.0 `family`, `origin`, `post_id`, `post_text` and `post_author_name`, the post a like or comment was on. See Raw signals. Changed in v1.52.0: `path_tags`, the tags on the account’s warm intros
`list_record_activity`	Changed in v1.28.0. An account’s, person’s or warm intro’s activity, including Clustr’s own actions and who made them. `entity: "path"` takes a warm intro’s opportunity id or `"<connector id>\|<prospect id>"` and lists who set its owner and status, and when
`create_saved_view`	New in v1.12.0. Save a filter as a view. Takes the encoded filter state and the member query your client already builds — this tool does not interpret filter syntax itself
`set_workflow_enabled`	New in v1.12.0. Turn a Workflow on or off
`set_account_tier`	New in v1.12.0. Set an account’s tier
`get_account_note`	New in v1.24.0. Read the note on an account, with who last edited it and when (`notes_updated_at`, `notes_updated_by_name`)
`set_account_note`	New in v1.24.0. Write the note on an account. It replaces the note there was, an empty note clears it, and it touches nothing else on the account
`clear_account_note`	New in v1.24.0. Remove the note on an account. Flagged destructive
`set_account_cs_owner`	New. Set an account’s CS owner, beside its owner. `push_to_crm` also writes it to the CRM company field mapped for the CS owner; the result’s `crmPushSummary` says what happened there
`set_account_owner`	Changed in v1.15.0. Set an account’s owner. `push_to_crm` also writes it to the matched CRM company; the result’s `crmPushSummary` says in one sentence what happened there, and the owner is saved in Clustr either way
`request_intro`	New in v1.12.0. Ask a connector to make an introduction
`restore_path`	New in v1.12.0. Bring back a dismissed path
`dismiss_account`	Changed in v1.24.0. Dismiss an account and every path there. Takes an optional `reason`, kept as the account’s `dismiss_reason`; a dismissal never writes the account’s note
`dismiss_path`	New in v1.12.0. Dismiss one path
`set_path_owner`	New in v1.33.0. Set one owner (List Owners) on many warm intros at once, or clear it with `owner_id: "unassigned"`. Takes `paths[]`, up to 25,000; notes and status are left alone
`set_connectors_owner`	New in v1.33.0. Set one owner (List Owners) on many connectors at once, as `set_connector_owner` does for one: their warm intros move with them. `owner_id: "unassigned"` clears the connectors’ owner and leaves their warm intros’ owners. Takes `connector_linkedin_ids[]`, up to 25,000
`set_connector_slots`	New in v1.33.0. Keep many connectors fresh at once (`in_slot: true`), or stop refreshing them (`in_slot: false`). Filling takes the free slots in the order given; the result says how many changed and whether slots ran out
`update_saved_view`	New in v1.13.0. Rename, refilter, or (un)share a view you own
`delete_saved_view`	New in v1.13.0. Delete a view you own
`create_workflow`	Changed in v1.14.0. Create a Workflow: a view, a trigger (enter, leave, or schedule), and one or more actions. The action kind `add_tag` (tag ids from `list_tag_groups`) replaced `set_status`, on accounts and on people
`update_workflow`	New in v1.13.0. Change a Workflow’s trigger or actions; only provided fields change
`delete_workflow`	New in v1.13.0. Delete a Workflow and its run history
`run_workflow`	New in v1.13.0. Run a Workflow now instead of waiting for its next pass
`restore_account`	New in v1.13.0. Bring back a dismissed account
`set_intro_status`	New in v1.13.0. Move an intro request from requested to sent, accepted, or declined
`create_tier`	New in v1.13.0. Add a tier to your workspace’s ranking
`update_tier`	New in v1.13.0. Rename a tier or move it in rank
`delete_tier`	New in v1.13.0. Delete a tier; every account carrying it keeps its other fields, just untiered
`bulk_set_account_state`	Changed in v1.15.0. Apply the same tier or owner change across a selection of accounts (tags go through `tag_records`). `push_to_crm` also writes the owner to each matched CRM company, and `crmPushSummary` reports how many were updated
`set_account_standing`	New in v1.33.0. Put many accounts on the target list (`action: target`), take them off it (`untarget`, rows a CRM list put there stay), add them to the client list (`client`), or exclude them everywhere (`exclude`, refused when it would take the exclusion list past 500). Takes `company_ids`, up to 25,000
`start_bulk_job`	New in v1.33.0. Change every record a table’s filters match in the background, for selections over 25,000 records, up to millions. Warm intros (`entity: "path"`) take `set_owner`, `add_tag`, `dismiss` and `restore`; `filters` is the list’s own query. Returns the job at once. Flagged destructive
`list_bulk_jobs`	New in v1.33.0. Your background changes, the ones running (`state: "active"`) and the ones finished in the last 7 days (`state: "recent"`), with status, progress and counts
`get_bulk_job`	New in v1.33.0. One background change: status, how many records it matched, went through and changed, and the reason in plain words when it failed
`cancel_bulk_job`	New in v1.33.0. Stop a background change. Changes already made stay made
`list_buyer_titles`	New in v1.34.0. Your buyer titles: the job titles of the people you sell to
`add_buyer_titles`	New in v1.34.0. Add buyer titles. Applies to everyone already imported within seconds
`delete_buyer_title`	New in v1.34.0. Remove one buyer title, by the id from `list_buyer_titles`
`list_excluded_titles`	New in v1.34.0. Your excluded titles: anyone whose job title or headline matches one is hidden
`add_excluded_titles`	New in v1.34.0. Add excluded titles. Nothing is deleted; removing the title brings the people back
`delete_excluded_title`	New in v1.34.0. Remove one excluded title, by the id from `list_excluded_titles`
`get_crm_cleaning_queue`	New in v1.37.0. The CRM Cleaning queue: the changes waiting for a decision, grouped by kind, with counts
`apply_crm_cleaning_group`	New in v1.37.0. Apply every waiting cleaning change of one kind. The changes are written to your CRM by a background worker. Only certain merges apply in bulk; Likely merges are accepted one at a time with `accept_crm_cleaning_change`. Flagged destructive
`dismiss_crm_cleaning_group`	New in v1.37.0. Dismiss every waiting cleaning change of one kind. By default Clustr also stops suggesting that kind; pass `suppress: false` to keep the suggestions
`list_crm_cleaning_findings`	New in v1.37.0. The findings report: problems Clustr cannot fix by itself, counted by kind and listed with a link to each record. Changed in v1.41.0: also duplicate deals, each with its records
`dismiss_crm_cleaning_findings`	New in v1.37.0. Dismiss every open finding of one kind; `suppress: true` stops reporting it
`merge_crm_duplicates`	New in v1.37.0. Merge one duplicate group from the findings report. HubSpot only. A merge is permanent in the CRM; revert within 30 days recreates the merged-away record as a new record with a new id. Changed in v1.41.0: also merges a group of duplicate deals; a deal merge cannot be reverted. Flagged destructive
`get_crm_completeness`	New in v1.37.0. How complete your CRM is, field by field, with a weekly trend. This is what the Overview shows
`get_crm_cleaning_summary`	New in v1.37.0. What CRM Cleaning did over the last few days: the weekly summary email’s content
`start_crm_cleaning_scan`	New in v1.39.0. Start a cleaning scan. Changes whose rule is set to auto (Do it) are written to your CRM, so check the settings first. Flagged destructive
`start_crm_enrichment`	New in v1.39.0. Fill missing company and contact fields, inside the monthly enrichment budget. Does nothing while enrichment is off
`list_crm_cleaning_runs`	New in v1.39.0. Cleaning runs, newest first: scheduled scans, scans started by hand, enrichment runs
`get_crm_cleaning_run`	New in v1.39.0. One run and its counts: what is waiting, applied, failed
`list_crm_cleaning_changes`	New in v1.39.0. The queue one change at a time, with before and after values and a link to each record
`accept_crm_cleaning_change`	New in v1.39.0. Apply one waiting change. It is written to your CRM by a background worker. Flagged destructive
`dismiss_crm_cleaning_change`	New in v1.39.0. Dismiss one waiting change or one finding. Nothing is written to your CRM
`find_crm_record_duplicates`	New in v1.41.0. The likely duplicates of one contact or company, with why they matched and whether the match is certain. Read-only
`list_crm_cleaning_history`	New in v1.39.0. Every change cleaning made or proposed, with whether it can still be reverted
`revert_crm_cleaning_change`	New in v1.39.0. Undo one applied change within 30 days. Phone repairs and deal merges cannot be reverted. A contact or company merge is permanent in the CRM; revert recreates the merged-away record as a new record with a new id. Flagged destructive
`retry_crm_cleaning_change`	New in v1.39.0. Try a failed or conflicting change again. Flagged destructive
`get_crm_cleaning_settings`	New in v1.39.0. The cleaning settings: each rule’s mode, the schedule, enrichment and the merge switch. Changed in v1.45.2: enrichment includes `when_different` (keep or suggest, per object type). Changed in v1.45.3: unset, contacts suggest. Changed in v1.45.5: `effective_when_different` gives the result with the defaults applied
`update_crm_cleaning_settings`	New in v1.39.0. Change the cleaning settings. Each rule is off, review (Ask me) or auto (Do it). Only the fields you pass change. Changed in v1.41.0: takes `realtime_duplicates`, the two data quality alerts, `primary_rules` and `match_rules`. Changed in v1.45.1: `duplicate_spike_alert` and `completeness_drop_alert` turn highlights in the weekly summary email on or off, not daily emails; `duplicate_spike_threshold` counts new groups in a week. Changed in v1.45.2: takes `enrichment_when_different` (`contact`, `company`: keep or suggest); enrichment fills blank fields only, and with suggest a different value waits for a person
`list_crm_cleaning_templates`	New in v1.39.0. The one-click cleaning recipes and the settings each applies
`get_crm_enrichment_budget`	New in v1.39.0. The monthly enrichment budget: the cap, what is used and when it resets. Changed in v1.41.0: also this month’s match rate. Changed in v1.45.2: `records_matched` counts the records the provider found, filled or not; adds `records_filled`, `records_already_complete`, `records_not_found` and `match_by_object`
`get_crm_cleaning_capabilities`	New in v1.39.0. What cleaning can find and fix on your CRM, detector by detector
`get_crm_health_score`	New in v1.39.0. The older CRM Health Score from the latest scan, with its trend. The Overview shows completeness instead (`get_crm_completeness`)
`revert_crm_cleaning_batch`	New in v1.40.0. Undo every applied change of one rule at once, optionally for one run, field or time window. Merges and phone repairs are left out. At most 1,000 changes per call; call again with `next_cursor` while `has_more` is true. Flagged destructive
`get_crm_cleaning_suggestions`	New in v1.40.0. The rules your team always accepts and could switch from Ask me to Do it, from the last 90 days of decisions
`export_crm_cleaning_history`	New in v1.41.0. The cleaning history as CSV text, with the same filters as `list_crm_cleaning_history`. At most `limit` rows (default 200, max 2,000) and about 60 KB of text, with a note when there are more
`export_crm_cleaning_findings`	New in v1.41.0. The open findings as CSV text, with the same filters as `list_crm_cleaning_findings`. At most `limit` rows (default 200, max 2,000) and about 60 KB of text, with a note when there are more
`get_crm_property_insights`	New in v1.41.0. Every property of your contacts, companies and deals with how many records fill it; custom properties with no data or rarely filled are flagged. HubSpot only. Read-only. Changed in v1.43.0: the count’s progress while it runs (`counted`, `toCount`, `etaSeconds`), `recordsWithoutValue` per property, and per object the fill bands and the trend of the latest 12 counts (`fillBands`, `history`, `previous`). Changed in v1.45.1: the automatic count runs monthly, at night
`refresh_crm_property_insights`	New in v1.41.0. Recount the property insights in the background, once an hour. Nothing is written to your CRM. Changed in v1.43.0: counting starts within a minute, whatever the hour, at one CRM search a second. Changed in v1.45.1: the automatic count it saves you waiting for is monthly
`list_crm_field_values`	New in v1.44.0. The spellings of one CRM field’s values grouped into clusters of the same value, with a suggested value to keep and the saved mappings. Without a field, the fields you can standardise. Read-only. Changed in v1.44.1: HubSpot dropdown labels next to option values, the dropdown’s options, and the mappings a scan did not apply, with how many records and why
`set_crm_value_mapping`	New in v1.44.0. Save a value mapping: every listed spelling of a field becomes one value. The Standardise values rule applies it, asking first by default
`delete_crm_value_mapping`	New in v1.44.0. Delete a value mapping. Rewrites already applied are not undone
`push_to_crm`	New in v1.13.0. Push an account or person to your connected CRM. Takes a `provider` argument and creates the record if it isn’t there yet, skips if it already is
`whats_new`	The MCP changelog: the latest 5 releases by default. `versions` (a number, or `all`) shows more, and `since` every release newer than a version you know. Changed in v1.44.3

New in v1.66.0. The default owner tools are gone.

REMOVED: get_record_settings and update_record_settings, deprecated in v1.64.0. A warm intro’s owner is the owner picked on it, else the connector’s owner, else the prospect’s account owner. See Who owns a warm intro.

New in v1.65.0. One tag field per record type.

CHANGED: every account, connector and warm intro has one tag field, Account tags, Connector tags and Intro tags, and each takes any number of tags.
CHANGED, breaking: create_tag takes subject_type instead of group_id. create_account_tag and start_bulk_job’s connector set_tag no longer take group_id.
RETIRED: create_tag_group, update_tag_group, delete_tag_group and reorder_tag_groups. A call says what to use instead.

New in v1.64.0. One owner on warm intros.

CHANGED: a warm intro’s owner is the owner picked on it, else the connector’s owner, else the prospect’s account owner (the Accounts sales owner). See Who owns a warm intro.
DEPRECATED: get_record_settings and update_record_settings. The default owner decides nothing, and update_record_settings changes nothing.

New in v1.63.0. Export the Customers table, and every table’s file holds every column it shows.

NEW: export_competitor_customers returns every competitor’s customers in one call, with the filters of list_competitor_customers. Each row adds people_you_can_reach, sales_owner and cs_owner. The REST route’s CSV is the Customers table’s file.
CHANGED: the export_competitor_accounts CSV adds standing, possible_customer_of, customer_of_confidence, lost_from, in_market, reading_label, buying_intent_label, likes, comments, first_touch, sales_owner and cs_owner. The export_competitor_interactions CSV adds what_happened, pitch, account_standing and rep_relationship.

New in v1.61.0. See which accounts show buying intent.

CHANGED: scoped list_competitor_accounts and export_competitor_accounts rows carry latest_intent: the newest post at the account, engaged by a competitor, that shows the person is hiring for the role, evaluating tools, migrating, running an RFP or starting a project, with the reason and who posted it. intent (any, or one intent) keeps those accounts only. The CSV adds buying_intent and buying_intent_reason.
CHANGED: list_competitor_interactions takes intent. The export_competitor_interactions CSV adds buying_intent and buying_intent_reason.

New in v1.60.0. Add the customers you know, and put competitor facts on your CRM companies.

NEW: add_competitor_customer adds a company you know is one of a competitor’s customers. It is saved as confirmed, stays listed when no evidence is found, and shows with sources ["user"].
CHANGED: a competitor website that was only partly read is read again on the next weekly run, retrying the pages that failed, and searched harder for customer pages when none were found.
NEW: get_competitor_crm_properties and set_competitor_crm_properties. Clustr can write the competitors a company is a customer of, how sure it is, and the date a competitor last engaged the company onto that company in your CRM. Each is off until you turn it on.
CHANGED: update_competitor takes title_groups, the roles followed at that competitor; [] goes back to the workspace job titles.
CHANGED: scoped list_competitor_accounts rows carry in_market, true when in the last 30 days two or more competitors engaged the account or three or more people there were engaged, and in_market_facts, the counts behind it. in_market=true keeps those accounts only.

New in v1.59.0. The team refresh follows the people it finds.

CHANGED: the monthly team refresh, and run_competitor_team_refresh, follow the joiners they find automatically while you have free slots; past your limit they are added paused, not followed. There is no review queue: pending is normally empty.
UNCHANGED: validate_competitor_joiners stays for compatibility and normally finds nobody waiting.

New in v1.58.0. Every tool that overwrites data already on your records asks first.

CHANGED: set_path_owner, set_connector_owner, set_connectors_owner, bulk_set_account_state, tag_records, set_connector_tags, start_bulk_job and apply_crm_cleaning_group work in two steps. The first call changes nothing and returns confirmationRequired, a summary and the counts of what it would overwrite; the same call with confirm: true, after you agree, applies it.
CHANGED: set_client_rule returns a preview (how many companies the saved and the new rule match) and saves nothing when a change could remove clients from your list; set_tag_group_crm_mappings warns before it lets a CRM own a tag field; set_crm_value_mapping asks first while Standardise values runs on Do it. Each applies with confirm: true.

New in v1.57.0. The default owner for warm intros can be read and changed from your assistant.

NEW: get_record_settings and update_record_settings. Changing the default owner reassigns the owners of your existing intros, so update_record_settings first returns a preview and saves nothing; it saves only when called again with confirm after you agree.
CHANGED (REST): PUT /api/tenant/record-settings takes confirm. A change that would move a stored owner answers confirmationRequired: true with its preview until it is sent with confirm: true.

New in v1.56.0. Lists are retired: saved views replace them.

REMOVED: create_list, update_list, delete_list, add_to_list and remove_from_list, and the add_to_list action on create_workflow and update_workflow. Every list was already copied to a saved view of the same name.
REMOVED (REST): /api/public/lists and its members routes.

New in v1.55.0. An import can say its warm intros are not all calculated yet.

NEW: get_job_status and list_jobs carry derive_incomplete: true, derive_incomplete_reason (stalled or backstop) and derive_note on a COMPLETED import whose derivation did not finish inside the run. The import is usable and the remaining intros appear over the next hours.

New in v1.54.0. Tags work the same way everywhere: every tag sits in a field.

NEW: reorder_tags and reorder_tag_groups; get_tag_group_crm_mappings and set_tag_group_crm_mappings.
NEW: create_account_tag takes group_id; list_tags and list_account_tags return each tag’s group_id.
CHANGED: set_account_tags and set_connector_tags replace only within the fields of the tags given. An empty list still clears everything. set_connector_tags takes several tags.
FIXED: creating an account or connector tag no longer starts a new General or Relationship field beside your own.

New in v1.53.0. Competitor collection is every two weeks.

CHANGED: each competitor is collected every two weeks. The schedule is still a weekly slot, Mondays at 11:00 Europe/Paris by default, and a competitor that ran in the last ten days skips it. get_competitor_monitoring_schedule still reads cadence weekly: it is the slot.
The team refresh (joiners, leavers and role changes) is still monthly.

New in v1.52.0. Filter on the tags of a related record, and account tag writes work again.

NEW: search_opportunities takes account_tag (the prospect’s account); list_accounts takes path_tag (accounts with at least one warm intro so tagged); list_connectors takes account_tag and path_tag; list_signals takes path_tags. Tags from one group match any; tags from different groups must all match.
FIXED: set_account_tags and tag_records on accounts answered “Something went wrong on our side” for every account. They write again.

New in v1.51.0. Competitor collection is weekly again.

CHANGED: every competitor is collected once a week, Mondays at 11:00 Europe/Paris by default. update_competitor_monitoring_schedule takes day_of_week (0 = Sunday) with hour, minute and timezone.
The team refresh (joiners, leavers and role changes) is still monthly, inside one of the weekly runs.

New in v1.49.0. The people behind competitor activity, and pitches.

NEW: list_competitor_customer_people: the contacts found at one competitor customer, and up to 20 buyers there your network reaches, each with the best way in.
NEW: scoped list_competitor_accounts and export_competitor_accounts rows carry people (up to 10 of the people engaged, with their newest interaction and its pitch fields) and pitches; both take order=standing, types and pitch on every scope. The CSV adds people_reached and pitches.
NEW: list_competitor_customers takes order=standing and owner; rows carry the CRM account owner and CS owner.
NEW: list_competitor_interactions takes account.
CHANGED: preview_competitors people carry group.
NEW (REST): List Competitor Customer People, and the same parameters and fields on the competitor accounts, customers and interactions endpoints.

New in v1.48.0. One tool per list, taking one item or many.

NEW: exclude_companies: exclude one company or up to 500 in one call. Each takes any of domain, linkedin_url, linkedin_numeric_id and company_id. Replaces exclude_company.
NEW: add_client_companies: add one client company or up to 1,000. Replaces add_client_company and import_client_companies.
NEW: add_owners: add one owner or up to 2,000. Replaces create_owner; import_owners still works as another name for it.
BREAKING: exclude_company, add_client_company, import_client_companies and create_owner no longer run. A call answers with the tool and arguments to use instead.
NEW (REST): the excluded companies API: GET, POST, POST /import and DELETE /{id} on /excluded-companies.

New in v1.47.0. Who your competitors are talking to, filtered and explained, and a say in who is followed when adding a competitor.

NEW: list_competitor_accounts and export_competitor_accounts with scope=all take in_crm, activity_only and owner (the CRM account owner’s email).
NEW: scoped rows carry people_engaged, latest (the newest interaction: who at which competitor did what on whose post) and the CRM owner and CS owner. The CSV adds people_engaged, latest and owner.
NEW: preview_competitors takes title_groups and lists the first page of people found; register_competitor takes title_groups and people, each followed or skipped. Skipped people are never added by a later search.

New in v1.46.0. One list of the accounts competitors are present at, and its export.

NEW: list_competitor_accounts with scope=all: the accounts competitors touch on either side plus every company that is a customer of a competitor you watch, even with no interaction. Narrow it with yours (your client list or not) and customer_of (any, or one competitor). Every scoped row carries customer_of and we_own.
NEW: export_competitor_accounts: the same list in one call, unpaged, up to 50,000 rows, with scope, period, competitor_ids, yours and customer_of.
NEW (REST): GET /competitor-accounts?scope=all and GET /competitor-accounts/export, as JSON or, with format=csv, as a CSV file.

New in v1.45.9. Target accounts say when their LinkedIn page could not be found. See List Target Accounts.

NEW: list_target_accounts rows carry linkedin_lookup: not_found when the LinkedIn company page given for the account could not be found after two lookups, a week apart. It may be a wrong or outdated URL. The account is then matched by its name when exactly one company carries that name.
CHANGED: an account whose LinkedIn page Clustr does not know yet is looked up in the background, at most twice per page.

New in v1.45.8. A new target account needs its LinkedIn company page. See Create Target Account.

CHANGED: create_target_accounts needs company_linkedin_url on every account, like linkedin.com/company/acme. An account without a valid one is not created and comes back in rejected with its index and the reason; the others are added. A call where every account is rejected fails with invalid_linkedin_url.
CHANGED: set_account_standing with target skips a company that has no LinkedIn company page and counts it in skipped.
CHANGED: a bound CRM list adds a company only with its LinkedIn company page, from the CRM record or from the company Clustr matched it to. See Target accounts from a CRM list.
Accounts already on the list are unchanged, with or without a URL.

New in v1.45.7. Totals that agree, and paging that works.

CHANGED: get_network_stats counts the warm intros the People list shows, the number search_opportunities totals with no filters, and the connectors list_connectors totals. It used to count dismissed, overlap-only and undisplayable paths, and people removed from the network.
FIXED: list_target_accounts takes page and page_size; list_jobs honours them.
FIXED: a warm intro’s note being set or cleared shows in list_record_activity for the path.

New in v1.45.5. get_crm_cleaning_settings returns effective_when_different: what enrichment does with a value your CRM already has, per object type, with the defaults applied (contacts suggest, companies keep).

New in v1.45.4. Competitor Attack alerts say whether their account is a target account or an other account. See Competitor alerts.

CHANGED: Attack covers your target accounts and every other non-customer account a competitor engages with. Each alert is labelled Target account or Other account in Slack and email, and carries account_scope (target or other) and account.is_target in the webhook.
CHANGED: get_competitor_alerts and update_competitor_alerts take attack_other_accounts: false keeps Attack alerts to target accounts only. Default true.

New in v1.45.3. CRM Cleaning fills a contact’s job title from the person’s current LinkedIn role at the contact’s own company, never from a role held elsewhere. See Enrichment.

CHANGED: enrichment_when_different defaults to suggest for contacts: a title or company LinkedIn gives differently waits in the queue and is not counted in the weekly email. Companies still default to keep.

New in v1.45.2. CRM Cleaning enrichment never overwrites by default, and its match rate counts what the provider found. See Enrichment.

CHANGED: enrichment fills blank fields only. update_crm_cleaning_settings takes enrichment_when_different (contact, company: keep, the default, or suggest); with suggest a different value is proposed as a change that always waits for a person, with a reason such as “Your CRM says X; LinkedIn now says Y”.
CHANGED: get_crm_enrichment_budget counts records_matched as the records the provider found, whether or not a field was filled, and adds records_filled, records_already_complete, records_not_found and match_by_object. Contacts not found are counted from October 2026.

New in v1.45.1. CRM Cleaning is calmer: one summary email a week, plus an immediate email only when changes fail to write to your CRM. See Completeness and the weekly email.

CHANGED: unusually many new duplicates and a field getting emptier are short highlights in the weekly summary, not daily emails. update_crm_cleaning_settings keeps duplicate_spike_alert, completeness_drop_alert and duplicate_spike_threshold (now new groups in a week) to turn them on or off.
CHANGED: property insights are recounted monthly instead of weekly, still at night. refresh_crm_property_insights still recounts on demand, once an hour.

New in v1.45.0. Competitor customers you can trust. See Their customers.

CHANGED: list_competitor_customers labels each customer Confirmed, Likely or Possible from the signals found on the competitor’s own website and posts, lists the contacts found and marks possible churn. Only Confirmed and Likely by default.
NEW: set_competitor_customer_renewal_date, get_competitor_customer_settings, update_competitor_customer_settings.
NEW: get_competitor_site, preview_competitor_website, recrawl_competitor_site. register_competitor and update_competitor take website_url.
CHANGED: lookup_competitor_customers_by_company returns Confirmed and Likely customers only, with the label.

New in v1.44.1. Dropdown labels in Standardise values. See Standardise values.

CHANGED: list_crm_field_values shows HubSpot dropdown labels next to option values (label, canonical_label, mapped_to_label, to_label), lists the dropdown’s options, and marks each mapping the latest full scan did not apply with skipped (reason, message, count, at).

New in v1.44.0. Standardise free-text values and filter the cleaning history. See Standardise values.

NEW: list_crm_field_values, set_crm_value_mapping, delete_crm_value_mapping.
CHANGED: list_crm_cleaning_history and export_crm_cleaning_history take who decided, rule, record, field and date filters.
CHANGED: states and provinces are written as their full name, like countries.

New in v1.43.0. Property insights show their progress and their trend. See Property insights.

CHANGED: get_crm_property_insights returns counted, toCount and etaSeconds while a count runs, recordsWithoutValue per property, and per object fillBands (how many properties are 0, 1-20, 20-50, 50-80 or 80-100% filled), history (the latest 12 counts) and previous (the count before the latest). The issue values are unchanged; the app calls them No data and Rarely filled.
CHANGED: counts run at one CRM search a second and pause for 15 minutes when HubSpot asks to slow down. The weekly count starts at night in your workspace’s time zone; refresh_crm_property_insights still starts at once.

New in v1.41.0. Better duplicates, data quality alerts and property insights. See CRM Cleaning from your assistant.

NEW: export_crm_cleaning_history: the change history as CSV text.
NEW: export_crm_cleaning_findings: the open findings as CSV text.
NEW: find_crm_record_duplicates: the likely duplicates of one record.
NEW: get_crm_property_insights and refresh_crm_property_insights: which HubSpot properties are filled, and which custom ones nobody uses.
CHANGED: get_crm_enrichment_budget also returns this month’s match rate: records sent, records with at least one field filled, and fills per field.
CHANGED: update_crm_cleaning_settings takes the two data quality alerts, realtime_duplicates, primary_rules (which record a merge keeps) and match_rules (your own duplicate rules).
CHANGED: duplicates are also found by phone number, new records are checked as soon as they are created, and duplicate deals are reported and can be merged one group at a time in HubSpot.

New in v1.40.0. Undo a whole rule, and know when a rule can run on its own. See CRM Cleaning from your assistant.

NEW: revert_crm_cleaning_batch: revert every applied change of one rule, for one run, field or time window, with the same checks as a single revert.
NEW: get_crm_cleaning_suggestions: the rules on Ask me that your team always accepts, ready to switch to Do it.
CHANGED: revert_crm_cleaning_change leaves a field someone edited in the CRM since as it is, and reports a conflict.

New in v1.39.0. The rest of CRM Cleaning from your assistant. See CRM Cleaning from your assistant.

NEW: start_crm_cleaning_scan, list_crm_cleaning_runs and get_crm_cleaning_run: start a scan and follow it. A scan writes the changes whose rule is set to auto (Do it).
NEW: start_crm_enrichment and get_crm_enrichment_budget: fill missing fields inside the monthly enrichment budget, and see what is left.
NEW: list_crm_cleaning_changes, accept_crm_cleaning_change and dismiss_crm_cleaning_change: work the queue one change at a time.
NEW: list_crm_cleaning_history, revert_crm_cleaning_change and retry_crm_cleaning_change: read what cleaning changed, undo a change within 30 days, retry a failed one.
NEW: get_crm_cleaning_settings, update_crm_cleaning_settings and list_crm_cleaning_templates: read and change each rule’s mode (off, review or auto), the schedule, enrichment and the merge switch.
NEW: get_crm_cleaning_capabilities and get_crm_health_score: what cleaning can fix on your CRM, and the health grade.

New in v1.38.0. Your client rule from your assistant, and competitor data on your accounts.

NEW: get_client_rule, set_client_rule, preview_client_rule, list_client_rule_fields, sync_client_rule and delete_client_rule: read and change who counts as a client in each connected CRM. See Client companies.
NEW: lookup_competitor_customers_by_company and resolve_linkedin_company.
CHANGED: list_accounts returns competitor_activity and takes the competitor filters and sorts; update_competitor_delivery_config takes routes; set_competitor_customer_verdict takes verdict: clear.

New in v1.37.0. CRM Cleaning from your assistant. See CRM Cleaning and CRM Cleaning from your assistant.

NEW: get_crm_cleaning_queue, apply_crm_cleaning_group and dismiss_crm_cleaning_group: read the queue and apply or dismiss a whole group of changes.
NEW: list_crm_cleaning_findings, dismiss_crm_cleaning_findings and merge_crm_duplicates: work the findings report, and merge duplicates on HubSpot.
NEW: get_crm_completeness and get_crm_cleaning_summary: field completeness with its weekly trend, and what cleaning did recently.

New in v1.34.0. Buyer titles and excluded titles. See Buyer titles.

NEW: list_buyer_titles, add_buyer_titles and delete_buyer_title: read and edit the job titles of the people you sell to.
NEW: list_excluded_titles, add_excluded_titles and delete_excluded_title: read and edit the job titles that hide a person everywhere.

New in v1.32.0. Import clients in one step.

NEW: import_connectors takes mark_companies_as_clients. Set it when the people you import are clients: once the import finishes, each connector’s current company joins your client list, so warm intros into those companies are hidden. Off by default.
CHANGED: import_connectors answers status: "started" with the job_id and says the import is running, not done.

New in v1.31.0. Slack channel access. See Slack channel access.

NEW: list_slack_channels, update_slack_channel_access and join_slack_channel: choose, channel by channel, whether Clustr may read a Slack channel, post to it, both or neither, and add Clustr to public channels.

New in v1.30.0. Every failed job says why.

NEW: get_job_status and list_jobs give a FAILED job a failure_reason, a plain-English sentence with what to do next (“Company not found on LinkedIn. Check the link points to the company’s own page and retry.”), and a stable reason_code such as company_not_found or timed_out. See Get Job Status.
CHANGED: get_account_research answers status: FAILED with the reason when the last run on the company failed. start_account_research says why a run was not started, and the magic link says a failed run failed instead of “processing”.
FIXED: an account research run that stopped while starting is recorded as FAILED with its reason. It used to stay PROCESSING.
CHANGED: list_signals rows name the post a like or a comment was on (post_text, post_author_name), so a like says what was liked. See Raw signals.
NEW: list_signals says what kind of signal each row is: family (network, activity, prospect) and origin (prospect when the person did it, network when one of your connectors engaged their post), and filters on both with families and origins.

New in v1.29.0. Raw signals: every interaction across your network, by date. See Raw signals.

NEW: list_signals: LinkedIn reactions, comments and posts, newest first or oldest first, filtered by since/until, types, the person’s account (industry, size, revenue, country, tier, owner, tags) and seniority. Answers “all the LinkedIn likes from the past month”.

New in v1.28.0. Activity history on every record. See Record Activity.

CHANGED: list_record_activity takes entity: "path", a warm intro, and lists who set its owner and status and when, note edits, dismissals and intro requests.
NEW: a connector’s activity says when they were added and by whom (connector_added). Changing an intro’s owner or status, and a CRM sync changing a record’s tags, now show with who did it.
Older history was filled in from the dates Clustr already had; those entries carry backfilled: true.

New in v1.27.0. See which contacts an import could not bring in, and why.

NEW: list_failed_import_items: the failed people of one import, each with a plain-English reason, a stable reasonCode, the raw error and retryable. Read it when get_job_status reports failures.
NEW (REST): GET /api/public/jobs/{id}/items. See List Failed Import Items.

New in v1.24.0. Notes on every record. See Notes.

NEW: get_person_note, set_person_note, clear_person_note: one shared note per person in your workspace, whether they are a connector, a prospect or both. Read it before requesting an intro through a connector.
NEW: get_account_note, set_account_note, clear_account_note. The account note could be written through the API and read by no tool. The answer says who last edited the note and when.
A warm intro’s note: get_opportunity reads it, with who last edited it and when (notes_updated_at, notes_updated_by_name), and update_opportunity writes it.
CHANGED: an empty note clears it, on every note. Any member can edit a note, and notes are not synced to a CRM.
CHANGED: dismiss_account takes a reason, stored as the account’s dismiss_reason. A dismissal never writes the note.

New in v1.15.0. Ask whats_new for this list from your own client.

NEW: create_tag_group, update_tag_group, delete_tag_group — tag groups for accounts, people and opportunities (subject_type path). A single-choice group holds at most one of its tags per record.
BREAKING: create_tag, update_tag and delete_tag now work on every tag, not only connector tags. create_tag requires group_id (from list_tag_groups; connector tags are in a person group) and takes closes; position is gone from create_tag and update_tag. Tag ids did not change, so update_tag and delete_tag on a connector tag work as before.
NEW: push_to_crm on set_account_owner, bulk_set_account_state and set_connector_owner — also write the owner to the matched CRM record. The result carries crmPush and crmPushSummary, a sentence saying what happened in the CRM. The owner is saved in Clustr either way.
CHANGED: update_opportunity sets the opportunity’s own owner with owner_id (from list_owners), and unassign_owner clears it. That is the owner the People table shows. set_connector_owner is unchanged: it sets the connector’s owner and moves every opportunity the connector has.
CHANGED: owner filters take owner ids from list_owners; an owner email still resolves. list_owners entries carry crmSeat when a CRM is connected.
CHANGED: the retired create_status, update_status and delete_status now point at create_tag, update_tag and delete_tag.

The status tools are gone, and the tag tools replace them. Calling list_opportunity_statuses, create_status, update_status, delete_status or set_account_status answers with the tool to use instead of “Unknown tool”, so a stale manifest recovers: read the vocabulary with list_tag_groups, write the vocabulary with create_tag, update_tag and delete_tag, and put a tag on a record with tag_records. A former status is a tag in a group usually still named Status, with the id it always had.

v1.13.0 tools are listed with their intent, not their final argument schema. They ship alongside a Workflow editor and Views UI still being built, so the exact argument names may still move before release — check whats_new or reconnect for the authoritative schema once your client fetches the tool list.

The fit grade is relative to your configuration, not to other accounts.fit_grade is the fraction of your CONFIGURED targeting criteria an account meets — your target list, your tier vocabulary, each dimension of your Company ICP — banded A+ (at or above 0.90) through D. A criterion you never set cannot lower a grade, and nothing earns an A+ for being the best of a weak list.An account with NO grade is not a low-graded account. It means the workspace has configured no targeting at all, so there is nothing to score against. D is the opposite: criteria exist and this account met none of them. Do not treat the two the same, and do not tell someone their accounts are poorly qualified when the real answer is that nobody has defined what a good account is yet — point them at Targeting instead.fit_reasons ships with the grade, so say WHY an account scored what it did rather than repeating the letter.

Heat is timing, fit is quality. Keep them apart. temperature says how active an account is right now, from its open CRM deal and its LinkedIn and CRM signals; fit_grade says how well it matches your targeting and does not move with activity. A cold A+ is a good account nobody is talking to; a burning D is activity at an account outside your targeting. last_signal_at, last_signal_source and signals_30d are the evidence behind the heat, so quote them rather than the band alone.

Tiers are your words, and a tier is set on the ACCOUNT. list_tiers returns whatever your workspace calls them — “A / B / C” for one customer, “Strategic / Growth / Watch” for the next — so never assume a name. An unknown one is rejected rather than created: a typo in a 500-row import would otherwise become a permanent tier nobody chose.A tier attaches to the company behind a target account, which is why it also shows on the Accounts and People tables in the app, and why an account that has not matched a company yet cannot carry one. When that happens on create_target_accounts the tier is queued, not lost: it is applied automatically as soon as the match happens. On set_target_account_tier, where you are acting on one specific row, it is a conflict instead, because there the right answer is to fix the account’s LinkedIn URL.

Removed in v1.35.0: pause_competitor. Competitors can no longer be paused: every registered competitor is monitored until delete_competitor removes it, which also frees its slot. update_competitor refuses status for the same reason.

Removed in v1.4.0: start_competitor_monitoring. register_competitor starts the new competitor’s first scan itself and tells you whether it started. To collect on demand, use run_competitor_monitoring, which goes through the same per-workspace throttle as the API. Use list_competitor_runs to see what is running.

Reading account research you already have

Account research runs once per company and stays readable. Starting a run uses a credit, but reading one is free, so your assistant should look before it starts anything:

list_account_research shows what already exists, including reports your team ran in the web app.
get_account_research returns the report itself: who the warm buyers are and who can introduce you to each. Your assistant can summarize it, rank the paths or draft the intro request from it.
start_account_research only runs when nothing exists yet. If research already exists, it hands back that report instead of starting a duplicate, and says so with already_exists: true. To run a company again on purpose, for example after a big hiring wave, ask for a refresh (refresh: true).

A report started from the web app, the API or MCP is the same report, so research run by a teammate in Clustr can be read straight from Claude, ChatGPT or Dust. If an assistant built on an older version keeps starting new runs, point it at get_account_research in its instructions, for example: “To read account research, call get_account_research. Only call start_account_research when it returns nothing.”

Owners are a directory, not a seat

An owner is the person accountable for a connector, and therefore for every opportunity that connector produces. In Clustr that is a directory entry scoped to your workspace, not a login, so it does not have to be one of your teammates. An investor, an advisor, a former colleague, or a friend can own an opportunity. That has three consequences worth knowing when you ask an assistant to assign work:

No email is sent. Adding an owner with add_owners never notifies that person. If you want them to know, tell them yourself.
No access is granted. An owner cannot sign in, see your network, or read the opportunity. The entry is a label on your data.
No seat is used. Adding owners does not change your billing, so there is no reason to be sparing with them.

Owners who are workspace members come back with is_member: true, so an assistant can tell the two apart. Deleting an owner never deletes opportunities: every opportunity that person held is simply left with no owner, and delete_owner reports how many so you can reassign them.

Who owns a warm intro

A warm intro’s owner is the first of:

The owner someone on your team picked for that intro (set_path_owner).
The connector’s owner (set_connector_owner).
The sales owner of the prospect’s account, the same owner the Accounts page shows. It follows your CRM as your owner sync direction says.

There is no workspace default owner any more.

Ask in plain language

You don’t need to know the tool names above to use them — describe what you want and your assistant picks the right ones. A few examples of the sentence and what it becomes:

You say	Tools it calls
”Tell me in Slack when a B-grade account gets a path over 80%“	`create_saved_view` (the filter), then `create_workflow` (Slack, on enter)
“What has Acme’s account been up to?”	`list_record_activity`
”Who took over the Jane Smith intro, and when did it move to Contacted?”	`search_opportunities` to find it, then `list_record_activity` with `entity: "path"`
”All the LinkedIn likes from the past month”	`list_signals` with `since`, `until` and `types: ["reaction"]`
”Turn off the hot accounts Workflow, I’m getting too many pings”	`list_workflows` to find it, then `set_workflow_enabled`
”Move that Jane Smith intro to Sent”	`set_intro_status`
”Add these five companies to our target accounts”	`create_target_accounts`
”We’re done with Globex, dismiss the whole account”	`dismiss_account`
”Actually bring Globex back”	`restore_account`
”Push Jane Smith to HubSpot”	`push_to_crm`
”What tiers do we have, and put Acme on our top one”	`list_tiers`, then `set_target_account_tier` or `set_account_tier`
”Mark that intro Contacted”	`list_tag_groups` to find the tag, then `tag_records`
”Give Acme to Marine, and update HubSpot too”	`list_owners`, then `set_account_owner` with `push_to_crm`

Because a view is the question and a Workflow is what happens when the answer changes, most “tell me when” requests become exactly one saved view and one Workflow — the same two things you’d build by hand in the app.

Example Prompts

Here are some things you can ask your AI assistant once Clustr is connected:

“Which A+ accounts have no warm path yet?”
“Tag every account we share an investor with, then list them”
“Run account research on Salesforce and give me the report link”
“What account research do we already have?”
“Show me the account research on Stripe. Who is the best first intro?”
“Summarize the warm paths into Datadog from our existing research”
“Find referral opportunities at Stripe”
“What opportunities do we have through john-doe on LinkedIn?”
“Show me my top opportunities by confidence score”
“Give me an overview of my network”
“Who are my connectors?”
“Check the status of my last import”
“Mark opportunity X as Contacted”
“What notes do we have on that Stripe opportunity?”
“Note on Patrick that he is an investor, and to ask Helene before requesting an intro through him”
“What does our note on Acme say, and who wrote it?”
“Add to the notes on opportunity X that Marie is making the intro next week”
“Assign that Stripe opportunity to sam@northstar.vc, he is our investor, create the owner if needed”
“Who owns the most opportunities right now?”
“Add Stripe as a target account”
“List all our target accounts”
“What tiers do we use?”
“Add Stripe and Datadog as target accounts at our top tier”
“Move Acme to Tier 2”
“Which target accounts have no tier yet?”
“Add Globex as a competitor to monitor”
“Which of my open-deal accounts did competitors touch this week?”
“Show me every critical signal from Globex’s founders and AEs in the last 30 days”
“Who at Northwind Trading have our competitors been talking to?”
“Log that comment to the CRM, then snooze the rest of that thread for a week”
“Show me the latest competitor report and which accounts are at risk”

Authentication

Clustr’s MCP server supports two auth methods:

OAuth (sign-in) — authorize in your browser; Clustr issues and refreshes tokens automatically. No key to manage. Recommended for Claude Code, Claude Desktop, and ChatGPT.
API key — the same key as the REST API, set via the CLUSTR_API_KEY environment variable. Best for headless or shared setups.

Troubleshooting

Issue	Solution
`CLUSTR_API_KEY environment variable is required`	Set the `CLUSTR_API_KEY` env var in your client config
Authentication errors	Verify your API key is correct and active in API settings
Tools not appearing	Restart your AI client after updating config. New tools need no reinstall and no re-authorization: the client refetches the list on reconnect
Rate limited	Wait a moment and retry. The error message includes retry timing.
Large responses truncated	Use `page` and `page_size` parameters to paginate results

Getting Started

Views & Workflows

Competitor Monitoring

Integrations

CRM Cleaning

Setup

Configure Claude Code

Other clients (Claude Desktop, ChatGPT, Cursor)

Available Tools

Reading account research you already have

Owners are a directory, not a seat

Who owns a warm intro

Ask in plain language

Example Prompts

Authentication

Troubleshooting

​Setup

​Configure Claude Code

​Other clients (Claude Desktop, ChatGPT, Cursor)

​Available Tools

​Reading account research you already have

​Owners are a directory, not a seat

​Who owns a warm intro

​Ask in plain language

​Example Prompts

​Authentication

​Troubleshooting

Setup

Configure Claude Code

Other clients (Claude Desktop, ChatGPT, Cursor)

Available Tools

Reading account research you already have

Owners are a directory, not a seat

Who owns a warm intro

Ask in plain language

Example Prompts

Authentication

Troubleshooting