HighLevel MCP Multi-Account Support for Claude
The HighLevel Model Context Protocol (MCP) server enables Claude AI assistants to interact directly with your CRM data across multiple sub-accounts. Agencies can now connect once and work seamlessly across any authorized location without reconnecting.
This guide covers the multi-account MCP endpoint, authentication methods, available tools, setup instructions for Claude clients, and security features that keep your data protected.
For more visit- LeadConnector MCP Server
What is HighLevel MCP for Claude?
The HighLevel Model Context Protocol (MCP) server is an integration that allows Claude AI assistants to interact directly with your HighLevel CRM data and operations. MCP is an open protocol that standardizes how applications provide context to large language models.
The Anthropic-specific endpoint (https://services.leadconnectorhq.com/mcp/anthropic/v2) provides Claude with a standardized way to query, automate, and orchestrate everything in your HighLevel account through natural language commands — without you or the model needing to know the internal API details.
This endpoint works with Claude.ai, Claude Code (CLI), and Claude Cowork, giving you flexibility across different Claude environments.
Key Benefits
The multi-account MCP endpoint transforms how agencies manage their HighLevel sub-accounts through Claude AI.
Multi-Account Connection Overview
The Anthropic MCP endpoint supports both individual and agency-level connections, each optimized for different use cases.
When connecting as an individual user, you authorize Claude to access a single sub-account. This is the default connection mode and works identically to previous MCP implementations — the connection remains tied to that one location for its lifetime.
Agency installations allow you to select multiple sub-accounts during the authorization flow. You can choose one sub-account, several specific locations, or all available sub-accounts under your agency. Claude can then work across only those authorized locations in the same conversation, switching between them as you direct.
During installation, agencies choose which sub-accounts to include. The connection then operates on exactly those sub-accounts and no others. Each request runs against a single sub-account — either the one you name in your prompt, or one Claude identifies through the list_locations tool when the context is ambiguous.
Available MCP Tools
Rather than exposing hundreds of individual operations as separate tools, the MCP server provides six unified tools that Claude uses to discover, inspect, and execute the full operation catalog. This compact toolset prevents overwhelming the model's context window while maintaining access to the entire HighLevel API surface.
Finds customer or business records by name, email, phone, tag, or similar criteria. This is typically the first tool Claude calls when you ask about existing data.
Retrieves the full details of one or more records returned by search. Use this to get complete information about a contact, opportunity, or other entity.
Discovers available operations by intent (list, create, update, delete, etc.). Claude calls this to find which operations match your request before executing anything.
Inspects an operation's required and optional inputs before running it. This ensures Claude knows what data to pass and in what format.
Runs one operation against the HighLevel API, subject to your OAuth scopes and built-in safety checks. All create, update, and delete actions flow through this tool.
Lists the sub-accounts this connection can access. Claude calls this when the target sub-account is ambiguous or when you ask to see available locations. For single-location connections, this simply returns the one authorized sub-account.
When you ask Claude to "Find the contact with email jane@example.com, add the tag 'vip-2026', and create a new opportunity worth $5,000," Claude uses search to find Jane, search_operations to discover the tag and opportunity operations, describe_operation to inspect their inputs, and execute_operation twice to complete both actions. You receive a single confirmation message.
Authentication Methods
The MCP endpoint supports two authentication methods. OAuth is recommended because it exposes a broader set of scopes and operations than Private Integration Tokens.
One-click sign-in through the HighLevel consent flow. When you authorize the connection, you review and approve the scopes on the consent screen. OAuth requires no manual token storage or rotation and provides access to the widest set of scopes.
You can review or revoke OAuth access at any time from your HighLevel account settings.
Create a token under Settings → Private Integrations, selecting the scopes you want the token to have. Pass the token in the Authorization header (Bearer pit-your-token) when connecting your MCP client.
A PIT offers a more limited set of scopes compared to OAuth, which may restrict the operations Claude can perform.
In both authentication methods, you choose which scopes to grant the integration. The operations Claude can perform are filtered by these scopes — the assistant can only access data and features your authorization permits.
How to Connect Claude Clients
The HighLevel MCP endpoint works with three Claude environments. Follow the setup instructions for your client below.
- Open Claude.ai and navigate to Settings → Connectors → Add custom connector.
- Set the server URL to
https://services.leadconnectorhq.com/mcp/anthropic/v2. - Click Connect and complete the HighLevel sign-in flow (sign in → pick sub-account(s) → approve scopes).
- Start a new chat. The HighLevel MCP tools are now available.
- Test the connection by asking: "Find the last 5 contacts I added in HighLevel."
Run this command in your terminal:
claude mcp add --transport http leadconnector https://services.leadconnectorhq.com/mcp/anthropic/v2Alternatively, add it to .mcp.json at your project root:
{
"mcpServers": {
"leadconnector": {
"type": "http",
"url": "https://services.leadconnectorhq.com/mcp/anthropic/v2"
}
}
}On first use, Claude Code opens a browser for HighLevel authorization. Verify the connection with claude mcp list.
- In Claude Cowork, open the connectors / integrations settings.
- Choose Add custom connector.
- Set the server URL to
https://services.leadconnectorhq.com/mcp/anthropic/v2. - Complete the HighLevel sign-in flow (sign in → pick sub-account(s) → approve).
Your Cowork agents can now use the HighLevel MCP tools.
Instead of OAuth, you can pass a Private Integration Token in the Authorization header (Bearer pit-your-token). Create the token under Settings → Private Integrations and select the desired scopes before connecting.
Switching Between Sub-Accounts
For agency connections authorized to access multiple sub-accounts, Claude can switch between locations instantly within the same conversation. Simply mention the sub-account name in your request.
You: "Show me the latest opportunities in Downtown Clinic."
Claude retrieves and displays the opportunities for Downtown Clinic.
You: "Now create a contact in Westside Dental."
Claude switches to Westside Dental and creates the contact — no reconnection required.
When the target sub-account is unclear or not mentioned, Claude calls the list_locations tool to show available locations and asks you to choose. This ensures operations always execute against the correct sub-account.
If you connected to only one sub-account (the default for individual users), list_locations simply returns that location. You don't need to name the sub-account in your requests — Claude always knows which one to use.
Security & Permissions
The MCP endpoint implements multiple security layers to ensure your HighLevel data remains protected.
Available Operations & Coverage
The Anthropic MCP endpoint provides access to 625 operations across 40 product domains. The exact operations available to your connection are filtered by the OAuth or Private Integration Token scopes you grant.
Coverage includes:
Use the search_operations tool at any time to discover the exact operations available to your grant. Claude can help you explore capabilities by domain or intent (list, create, update, delete, etc.).
Frequently Asked Questions
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article