How to Use the HighLevel MCP Server

Modified on: Mon, 14 Jul, 2025 at 3:17 PM


This article will show you how to leverage the new GoHighLevel MCP (Model Context Protocol) server to empower your AI agents and copilots to access GoHighLevel data and tools seamlessly — all via a secure, standardized HTTP protocol.

TABLE OF CONTENTS


What is the MCP Server?

The GoHighLevel MCP server is a standardized, secure protocol that allows AI agents to read and write data in GoHighLevel without needing SDKs or deep custom integrations. You can now give agents access to critical tools like Calendar, Contacts, Conversations, Opportunities, and Payments through a single, unified interface.

Key Benefits of Using the MCP Server

Connecting your AI agents to GoHighLevel has never been easier or safer.

  • Standardized Access: Query and update data through a unified protocol.

  • Fast Integration: No SDKs required; works with any HTTP-supporting client.

  • Secure: Uses private integration tokens (PITs) for granular, permission-based access.

  • Scalable: Designed for future expansion to 250+ tools.

  • Flexible: Supports popular clients like Cursor, Windsurf, and OpenAI Playground.

Top 21 Tools Available

Below is the initial set of tools your AI agents can use today:

#

Tool

Endpoint

Description

1

Get Calendar Events

calendars_get-calendar-events

Get calendar events using userId, groupId, or calendarId.

2

Get Appointment Notes

calendars_get-appointment-notes

Retrieve notes for a specific appointment.

3

Get All Tasks

contacts_get-all-tasks

Retrieve all tasks for a contact.

4

Add Tags

contacts_add-tags

Add tags to a contact.

5

Remove Tags

contacts_remove-tags

Remove tags from a contact.

6

Get Contact

contacts_get-contact

Fetch contact details.

7

Update Contact

contacts_update-contact

Update a contact.

8

Upsert Contact

contacts_upsert-contact

Update or create a contact.

9

Create Contact

contacts_create-contact

Create a new contact.

10

Get Contacts

contacts_get-contacts

Fetch all contacts.

11

Search Conversation

conversations_search-conversation

Search/filter/sort conversations.

12

Get Messages

conversations_get-messages

Retrieve messages by conversation ID.

13

Send a New Message

conversations_send-a-new-message

Send a message to a conversation thread.

14

Get Location

locations_get-location

Get location details by ID.

15

Get Custom Fields

locations_get-custom-fields

Retrieve custom field definitions for a location.

16

Search Opportunity

opportunities_search-opportunity

Search for opportunities by criteria.

17

Get Pipelines

opportunities_get-pipelines

Fetch all opportunity pipelines.

18

Get Opportunity

opportunities_get-opportunity

Fetch opportunity details by ID.

19

Update Opportunity

opportunities_update-opportunity

Update opportunity details.

20

Get Order by ID

payments_get-order-by-id

Retrieve payment order details.

21

List Transactions

payments_list-transactions

Fetch paginated list of transactions.

How To Set Up the MCP Server

Step 1: Configure Your Agent/Client

Add the MCP endpoint and headers to your agent configuration:

json

CopyEdit

{

  "mcpServers": {

    "prod-ghl-mcp": {

      "url": "https://services.leadconnectorhq.com/mcp/",

      "headers": {

        "Authorization": "Bearer <your-token>",

        "locationId": "<sub-account-id>"

      }

    }

  }

}


Step 2: Get Your Private Integration Token (PIT)

  1. Go to Settings → Private Integrations in your GoHighLevel location.

  2. Click Create New Integration.

  3. Select the required scopes (listed below).

  4. Copy and securely store your token.

Required Scopes

Make sure your PIT includes:

  • View/Edit Contacts

  • View/Edit Conversations

  • View/Edit Conversation Messages

  • View/Edit Opportunities

  • View Calendars & Calendar Events

  • View Locations

  • View Payment Orders & Transactions

  • View Custom Fields

  • View Forms

Step 3: Start Building!

Your agents can now call tools using natural language instructions, seamlessly translated into secure API calls via MCP.

Example Tool Call

Python example:

python

CopyEdit

import requests


headers = {

    "Authorization": "Bearer YOUR_PIT_TOKEN",

    "locationId": "YOUR_LOCATION_ID"

}


data = {

    "tool": "contacts_get-contact",

    "input": {

        "contactId": "abc123"

    }

}


response = requests.post(

    "https://services.leadconnectorhq.com/mcp/",

    headers=headers,

    json=data

)


print(response.json())


Roadmap & Vision

The current MCP release is just the beginning. Our roadmap includes:

  • Individual MCP servers dedicated to specific services for targeted workflows.

  • npx package to support clients without native HTTP support (e.g., Claude Desktop).

  • OAuth support for advanced authorization flows.

  • Expansion to 250+ tools to create a true unified orchestrator layer for AI agents.

  • Enable complete automation of business processes across all GoHighLevel modules.


Frequently Asked Questions

Q: Can I use this with OpenAI Playground or Claude?
 Yes! Any client supporting HTTP requests can integrate with MCP.

Q: Do I need to install an SDK?
 No SDK is required — MCP uses a standard HTTP protocol.

Q: Is my data secure?
 Yes. Data access is fully scoped via Private Integration Tokens and secured through HTTPS.

Q: Can I restrict what tools an agent can use?
 Yes. You can control access by limiting scopes in your PIT.

Q: What if I need a new tool or endpoint?
 Reach out with feedback! We're actively expanding the available toolset.

Next Steps

  • Create your Private Integration Token and configure scopes.

  • Update your AI agent or client configuration with the MCP endpoint.

  • Start building agent-driven automations and integrations today!

  • Provide feedback to help shape future expansions and new tool support.



Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select at least one of the reasons
CAPTCHA verification is required.

Feedback sent

We appreciate your effort and will try to fix the article