GoHighLevel MCP Claude Setup: The Complete 2026 Guide

Why You Need the GoHighLevel MCP Claude Setup

Connecting Claude to your CRM changes how you manage client data. The Model Context Protocol (MCP) acts as a universal translator between AI models and external platforms. You no longer need complex Zapier webhooks or brittle API scripts to make Claude read your database. The official LeadConnector MCP server provides direct, secure endpoints for contacts, pipelines, calendars, and content publishing.

Once connected, you can type plain English commands to execute complex CRM tasks. Claude can read your pipelines, identify stuck leads, draft personalized SMS follow-ups, and even publish blog posts directly to your GoHighLevel sites. If you are wondering Is GoHighLevel Worth It in 2026? A Practitioner's Comparison, this native AI extensibility is a major reason why top agencies refuse to switch platforms.

Prerequisites for the Integration

Before starting the configuration, you must gather a few specific assets. You need agency admin or sub-account admin privileges to proceed. Private integrations live strictly at the sub-account level, meaning you cannot configure this globally across your entire agency dashboard.

You also need a paid Claude account using either the Claude Desktop application or the Claude Code command-line interface. The web version of Claude does not currently support custom local MCP servers. Set aside 15 minutes to complete the token generation and connection testing.

Step 1: Generate Your Private Integration Token

Your AI agent requires a secure key to access your CRM data. Navigate to Settings > Private Integrations in your target sub-account. Click the Create New Integration button in the top right corner. Name your integration clearly, such as "Claude MCP Agent", so your team knows exactly what this token does.

You must select specific scopes to grant access. Do not select every available scope. Over-permissioning creates unnecessary security risks. Start with read-only scopes like "Contacts Read" and "Opportunities Read" before granting write access. Click Create Integration and copy your token immediately. HighLevel will not show this token again, so paste it into a secure password manager.

Step 2: Locate Your Sub-Account Location ID

Your AI agent needs to know exactly which sub-account database to target. Go to Settings > Business Profile within the same sub-account. Look for the Location ID field under your core business information.

Copy this string of characters and save it alongside your integration token. While recent updates to the HighLevel API have made the Location ID header optional in some contexts, providing it explicitly in your connector configuration prevents cross-account data errors. For official documentation on multi-account routing, review the HighLevel MCP Multi-Account Support changelog.

Step 3: Configure the MCP Server in Claude Desktop

Open your Claude Desktop application to finalize the bridge. Navigate to Settings > Connectors > Add custom connector. You will need to input the official HighLevel MCP endpoint URL and your authentication headers.

Use https://services.leadconnectorhq.com/mcp/ for the server address. You must add two specific headers for authentication. First, add Authorization with the exact value Bearer YOUR_TOKEN (ensure you include the word Bearer and a single space). Second, add locationId with your copied Location ID string. Save the configuration and wait for Claude to confirm the connection.

Step 4: Alternative Setup Using Claude Code CLI

If you prefer working in the terminal, you can configure the server using Claude Code. Open your command line interface and ensure you have the latest version of Claude Code installed. You will use a single command to inject the transport URL and headers.

Run the following command, replacing the placeholder values with your actual credentials:

claude mcp add --transport http ghl-mcp https://services.leadconnectorhq.com/mcp/ -H "Authorization: Bearer YOUR_PIT_TOKEN" -H "locationId: YOUR_LOCATION_ID"

Hit enter to execute. You should see a success message stating the GHL MCP has been added. You can verify the active tools by typing /mcp directly into your Claude Code terminal.

Testing Your New AI Connection

Once connected, you must verify that Claude can actually read your CRM data. Open a new chat in Claude Desktop or your terminal. Type a simple discovery command like "Connect to GHL and list my pipeline names." Claude will execute the tool, fetch the API response, and return your pipeline data in plain text.

If you receive a 401 Unauthorized error, check your Bearer token formatting. The most common mistake is forgetting the space between "Bearer" and the token string. If Claude says it lacks permission to view pipelines, you need to return to GoHighLevel, edit your Private Integration, and add the "Opportunities Read" scope. For a deeper dive into the initial connection mechanics, read How to Connect Claude to GoHighLevel (2026 MCP Guide).

Required GHL GOAT Value Block: Pipeline Analysis Prompt

Once your MCP server is live, you need structured prompts to get reliable results. Use this exact directive to have Claude analyze your pipelines and draft personalized follow-ups without accidentally sending messages prematurely.

You are an expert CRM manager. Connect to my GoHighLevel account using the MCP tool.
First, pull all opportunities in the "Lead Management" pipeline.
Second, identify any opportunity that has been in the "New Lead" stage for more than 48 hours.
Third, draft a highly personalized SMS follow-up for each stuck lead based on their custom fields and past conversation history.
Do not send the SMS yet. Present the drafts to me in a markdown table for review.

For more copy-paste AI directives, the full SOP with screenshots lives free in the GHL GOAT library at Guided AI Directives.

Real Agency Use Cases for Claude MCP

The true value of this setup lies in bulk data operations. Consider database reactivation campaigns. You can instruct Claude to find all contacts tagged "Cold Lead" who have not replied in 90 days. Claude can then analyze their previous conversation history and draft unique, context-aware reactivation emails for each contact.

Content operations are another massive use case. The official server now includes tools for blog publishing and social media scheduling. You can paste a raw document into Claude and say, "Format this as a blog post, check if the slug 'mcp-guide' is available, and publish it to my GoHighLevel blog as a draft." Claude handles the formatting, the API call, and the draft creation in seconds.

Troubleshooting Common MCP Errors

Sometimes your AI agent will fail to execute a command despite a successful initial connection. First, verify your token has not expired. You can rotate tokens by going back to Settings > Private Integrations. Second, ensure your prompt is highly specific. If Claude tries to update a contact without passing a valid contact ID, the API call will fail with a 400 Bad Request error.

If you are building complex automations and things stall, review our GoHighLevel Workflow Not Triggering: 2026 Troubleshooting Guide to ensure your internal CRM logic is sound before blaming the AI. Often, a missing tag or a misconfigured pipeline stage is the actual culprit. You can also reference the official HighLevel MCP Server Setup Guide for endpoint-specific error codes.

Security Best Practices for API Tokens

Treat your Private Integration Token like a master password to your client's business. Never share a token across different sub-accounts. Always use one dedicated token per client location to maintain strict data isolation.

Rotate your tokens every 90 days to maintain strict security hygiene. When testing new write commands, always use a sandbox sub-account first. An AI agent can accidentally overwrite 100 contact records in seconds if your prompt is poorly phrased. Always instruct Claude to "show me the draft first" before executing bulk actions, sending emails, or publishing live content.