Abency
Run a marketing agency's social, SEO, ads, WordPress and email from an AI agent.
Read leads, contacts, opportunities and activities in Close CRM, and optionally write or delete them.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The Close MCP server connects Close CRM to Claude, ChatGPT, Cursor and any MCP-compatible agent. It reads leads, contacts, opportunities and activities, and can create, update or delete them. You grant one of three OAuth scopes at sign-in — read-only, safe writes, or writes including deletion — so the consent screen decides how much damage is possible.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
mcp.read unless you need writes.mcp.read scope at this time.perform_search is listed but undocumented. It appears in the directory snapshot and in no Close documentation we could find.RateLimit header with limit, remaining and reset; Close tells callers to sleep for the rate_reset value. Close publishes no MCP-specific figure, so the applicable numbers depend on which endpoint group a tool hits.No. Close states it does not currently support SSE, and asks teams that need it to email support@close.com with their use case and MCP client details. The server is Streamable HTTP only, which every current MCP client supports. Close's own n8n setup instructions specify HTTP Streamable as the transport for the same reason.
Close publishes the full list on its developer site, grouped by the three scopes. We counted 107 documented tools on 2026-08-18: 57 under `mcp.read`, 16 more under `mcp.write_safe` and 34 more under `mcp.write_destructive`. Anthropic's directory snapshot dated 2026-08-16 names only 52 of them, so the listing is a subset.
Yes. Close documents adding one custom connector per organization, each with a distinct name such as `close-mcp-org-name`, authenticating with either OAuth or the `Close-API-Key` and `Close-Scope` headers. Because Claude rejects two connectors sharing a URL, Close suggests appending an incrementing query string to the endpoint for each additional connection.
.md variant Close serves). The documentation URL in Anthropic's directory, https://help.close.com/v1/docs/en/mcp-server, HTTP 308-redirects here; this is the current canonical path. Source for the endpoint, the three-scope header model, the client setup steps, the paid-Claude-plan requirement, the n8n read-only note and the two FAQ answers about SSE and the tool list. · retrieved 2026-08-18.md, ~29KB). The scope-to-tool mapping: 107 tools under three explicit scope headings, with Close's own description of each. Source for every per-tool behaviour claim on this page, including delete_lead, enrich_field and schedule_voice_agent_call. · retrieved 2026-08-18.md). Source for the DCR statement, the per-scope one-line definitions, the multi-organization workaround and the third FAQ answer. · retrieved 2026-08-18initialize against https://mcp.close.com/mcp — HTTP 401 with WWW-Authenticate: Bearer error="invalid_token", error_description="Authentication required", resource_metadata="https://mcp.close.com/.well-known/oauth-protected-resource", retrieved 2026-08-18. This challenge is what establishes the OAuth posture. We never authenticated and never called a tool. · retrieved 2026-08-18scopes_supported of mcp.read, mcp.write_destructive, mcp.write_safe and offline_access, and names https://api.close.com/ as the authorization server. The /mcp-suffixed path returns the identical document. · retrieved 2026-08-18registration_endpoint (dynamic client registration), code_challenge_methods_supported of S256 (PKCE), and the authorization_code and refresh_token grants. It publishes no scopes_supported field, so the scope claim on this page rests on the protected-resource descriptor, which RFC 9728 makes authoritative for this resource. · retrieved 2026-08-18.md). Source for per-endpoint-group enforcement, the per-organization limit being 3× the per-key limit, the RateLimit header fields and the rate_reset guidance. · retrieved 2026-08-18.md). Source for the four predefined roles, the Scale-plan gate on custom roles, and the list of higher-risk actions — bulk import, bulk email, export, merge, delete leads — that we then checked against both tool lists. · retrieved 2026-08-18.md). Source for creating a key at Settings → Developer → API Keys, the one-time display, and the one-key-per-integration recommendation. · retrieved 2026-08-18partner tier, endpoint, transport and category above are read from our committed directory snapshot dated 2026-08-16. The directory page itself blocks automated fetches, so we did not retrieve it directly. · retrieved 2026-08-16help.close.com/robots.txt and developer.close.com/robots.txt — both retrieved 2026-08-18, both Allow: / with only /api/fern-docs/ disallowed. developer.close.com additionally carries Content-Signal: ai-train=yes, search=yes, ai-input=yes — an explicit opt-in to AI synthesis, which is the basis on which this page restates Close's documentation. Both hosts publish an llms.txt index and serve clean Markdown by appending .md to any page URL. · retrieved 2026-08-18privacy_policy. The page's own title is "Close CRM Security & Privacy" and it covers SOC 2 Type 2, GDPR and encryption while *linking* to the privacy policy rather than being it — so it is a combined security-and-privacy overview, not the policy document itself. · retrieved 2026-08-18Connect Close once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.