401 tools covering the full GoHighLevel API v3 — contacts, conversations, calendars, pipelines, invoices, knowledge bases, ad publishing (Facebook/Google/LinkedIn), social planner, and more — available to Claude, Cursor, Windsurf, or any MCP-compatible AI agent.
Run the server on demand via npx. No cloning, no compilation. Paste two env vars into your AI client config and you're done.
Contacts, conversations, calendars, opportunities, email marketing, funnels, phone numbers, payments, social planner, ad publishing (Facebook/Google/LinkedIn), Agent Studio, Voice AI, and a full knowledge base / web crawler suite.
Describe what you need — "move all leads untouched for 30 days to Lost" — and the AI handles the API calls. You stop writing GHL automation chains.
Discover, crawl, and train your GHL knowledge bases from natural language. Manage FAQs and ingest website content without touching the GHL UI.
Stdio transport — the MCP standard. Claude Desktop, Claude Code, Cursor, Windsurf, or any client that accepts a command + env config.
Detailed error messages surface GHL's 422 field validation and 429 rate limit errors with actionable guidance. No silent failures.
npx @nerdsnipe-inc/ghl-mcp-serverpnpm dlx @nerdsnipe-inc/ghl-mcp-serverbunx @nerdsnipe-inc/ghl-mcp-servernpm install -g @nerdsnipe-inc/ghl-mcp-serverRequires Node.js 18 or later. You will also need a GHL Private Integration Token and your sub-account Location ID — both take under two minutes to create in GoHighLevel Settings.
That's the entire configuration surface. Your Private Integration Token scopes which GHL capabilities the AI can access — you control permissions in GHL Settings, not in the server.
Workflow IDs, pipeline IDs, calendar IDs, and all other resource identifiers are passed as tool parameters at runtime — the server stays generic across any GHL account.
v3. The server was fully migrated from v2 to v3 in a 2.0.0 release that also added complete coverage of every practically-buildable v3 resource and a full audit fixing 79 request-schema mismatches against the real OpenAPI specs.
Resources that require agency-level or Marketplace-app-developer credentials rather than a location-scoped Private Integration Token: saas, marketplace, companies, snapshots, and oauth. Open an issue if you have the right credential type and need one of these.
Yes. The server is generic — it uses your Location ID and PIT token to scope requests to your specific sub-account. One config per sub-account.
Yes — npx runs it on demand without installing. Add npx @nerdsnipe-inc/ghl-mcp-server as the command in your MCP client config and it just works.
Clone the repo, run npm install, then point your MCP client at the source via npx tsx /absolute/path/to/src/index.ts. No build step needed for development.
GHL enforces 100 requests per 10 seconds. The server surfaces 429 errors with a clear message. For bulk operations, add delays between calls or process in batches.