The short answer
Yes. HighLevel launched its Model Context Protocol (MCP) server on July 30, 2025. It lets MCP-compatible AI clients such as Cursor, Windsurf, OpenAI Playground and Claude read and change data in a sub-account, including contacts, conversations, calendars, opportunities and payments. The basic setup uses the endpoint services.leadconnectorhq.com/mcp/, a Private Integration Token and the sub-account's location ID.
Connecting with a Private Integration Token
In the sub-account, open Settings, Private Integrations, create an integration, choose the scopes the AI should have and copy the token. Add the MCP endpoint to your client's configuration with two headers: an Authorization header carrying the token and a locationId header. The tools a client can see depend on those scopes, so an assistant given only view scopes cannot change records. HighLevel's guide says no SDK is needed because MCP runs over standard HTTP, and it lists OAuth for this endpoint as planned for a future release.
- Contacts: search, create, update and upsert, tags and tasks
- Conversations: search threads, read messages, send a new message
- Calendars and opportunities: events, appointment notes, pipelines and deal updates
- Payments: order details and transaction lists
- Content: blog posts, email templates and social media posts
OAuth endpoints for Claude and ChatGPT
HighLevel's developer docs list separate OAuth endpoints per client: services.leadconnectorhq.com/mcp/anthropic/v2 for Claude, including Claude.ai and Claude Code, and services.leadconnectorhq.com/mcp/openai/v2/ for ChatGPT, Codex and other OpenAI clients. You sign in, pick the sub-accounts to expose and approve scopes, with no token to create or rotate. An agency can authorize several sub-accounts in one connection and name the client in a prompt to switch between them, while each request still runs against a single sub-account. A small set of general tools lets the assistant discover and run hundreds of operations.
MCP inside HighLevel's own AI agents
The protocol also runs the other way. HighLevel's help center documents connecting external MCP servers to AI agents in workflows, so an agent can pull from web search, browser automation, databases or custom APIs, and there are guides for MCP connectors in Voice AI agents and managed agents. The setup guide for HighLevel's own MCP server does not list a separate fee. Because an AI client with edit scopes can send messages and change records, grant the narrowest scopes that do the job and test on a sandbox sub-account before touching client data.
More answers in HighLevel questions. For costs, see GoHighLevel pricing or run the cost calculator.