How to connect the GoHighLevel MCP server to n8n

GoHighLevel now has an official MCP endpoint, which means an n8n AI Agent can read contacts, check calendars and write opportunities without you building an HTTP Request node for every single endpoint. Setup takes about ten minutes. Making it safe takes a little longer, and that part matters more.

What this setup does

An MCP (Model Context Protocol) server exposes GoHighLevel’s actions, like reading a contact, creating an opportunity, listing calendar events or sending a message, as tools a language model can call. In n8n, the MCP Client Tool node connects to GHL’s server, pulls in that tool list and hands it to an AI Agent node. You give the agent an instruction in plain English, it picks the right GHL action, and n8n runs it.

So instead of a rigid workflow with one branch per action, you get an agent that can use whatever GHL operations the server offers. You still keep n8n’s error handling, logging and scheduling around it.

What you need

• A GoHighLevel sub-account (Location) with admin access
• n8n 1.88 or newer, cloud or self-hosted, for the MCP Client Tool and AI Agent nodes
• An LLM credential in n8n: OpenAI, Anthropic or Gemini
• Ten minutes to create a Private Integration token, which is how the MCP server authenticates you

Step 1: Find the GHL MCP server

There’s nothing to deploy. The endpoint is LeadConnector’s, at https://services.leadconnectorhq.com/mcp/. It authenticates every request with your Private Integration token and the Location it should act on. All you need is the right credentials.

Step 2: Create the Private Integration token

• In GHL go to Settings, then Private Integrations, and create a new integration at the sub-account level
• Grant only the scopes the agent will actually use, for example contacts.readonly, contacts.write, calendars.readonly, calendars/events.write, conversations/message.write
• Copy the token straight away (GHL shows it once) and note your Location ID from Settings, Business Profile

Wrong scopes are the most common failure. The connection opens fine, then a specific tool call comes back 401 or 403. If that happens, the same rules from my GHL 401 fix apply.

Step 3: Point the MCP Client Tool at the server

• Add an MCP Client Tool node, or attach it as a tool under an AI Agent node
• Set the server endpoint to https://services.leadconnectorhq.com/mcp/
• Add the headers the server expects:

Authorization: Bearer pit-xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
locationId: YOUR_LOCATION_ID
Version: 2021-07-28

Keep the token in an n8n Header Auth credential rather than pasting it inline, so it doesn’t end up in your workflow JSON. One space after Bearer and no trailing newline. A malformed header shows up as Invalid JWT.

Step 4: Run one real tool call

• Attach an AI Agent node with your LLM credential and connect the MCP Client Tool as its tool source
• The agent discovers the GHL tools on its own: contacts, calendars, opportunities, conversations
• Give it something concrete, like “Look up the contact with this email and tell me their pipeline stage”
• If that works, your token, scopes, Location and transport are all lined up, and you can build on top

Step 5: Guardrails before production

An agent that can write to a client’s CRM needs the same discipline as any other automation. I add these three before it touches live data.

• Scope the token tightly. If the agent only reads and books, don’t give it delete or opportunity-write. The server can only expose what the token allows, so the token is your blast radius.
• Dedupe writes. An agent that runs again on a retry can create a second contact or a duplicate opportunity. Key writes on a stable ID and drop repeats, the same way my production reliability checklist handles webhook retries.
• Log every tool call. Write the chosen tool, its arguments and the result to a table. When the agent does something odd, that log is the only way to find out why.

Things that will bite you

Missing locationId header. The server needs to know which sub-account to act on. Without it the connection can still open, but tool calls fail or hit the wrong Location. Set it explicitly.

Get Basit Qureshi’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

Agency token on a Location server. The endpoint works at Location level. An agency-level token authenticates but isn’t valid for Location tool calls, so create the integration at the sub-account level.

Transport mismatch. If n8n says it can’t list tools, the endpoint or transport is off. Use the exact /mcp/ path with the trailing slash and n8n’s default MCP transport. A wrong path returns an empty tool list, and the agent then decides it has no tools.

Vague prompts. Tell an agent to “clean up my CRM” and it will happily delete things if the token lets it. Keep instructions specific and keep the token scoped to match.

Where to take it next

• Trigger the agent from GHL events using the GHL to n8n webhook guide
• Add a second MCP server, like Supabase or Slack, so one agent can work across CRM and database
• Run it on a schedule for nightly enrichment or pipeline clean-up

The full guide and the importable workflow are on basitq.cloud, along with 4,900+ workflow templates.