MCP server for AI agents

The MCP server lets outside AI assistants work with your HLB account through the Model Context Protocol, a standard way for AI apps to find and use tools. From plain-language requests, a connected assistant can retrieve records, update supported data, send messages, review opportunities, check calendar information and take other actions you approve.

Compatible clients include developer editors such as Cursor and Windsurf, Claude environments that support MCP, the OpenAI Playground and custom applications that can connect to a remote MCP server.

To connect a client:

• Create a Private Integration token under Settings, then Private Integrations, for the account you want to connect, and choose only the scopes the AI needs.
• In your AI client, add the HLB MCP endpoint, the token and your location ID.
• Name the server however you like. The name in your client configuration is only a local alias.

Scopes control which resources the assistant can read or change, so it can only use the tools you grant. Some clients, such as Muse, sign in with OAuth instead. You review the requested permissions before granting access, and Muse can then reach more than 550 operations across 38 CRM areas through four unified tools, depending on the scopes you approve.

MCP also works in the other direction. HLB managed agents and Ask AI can connect to external apps such as Google Calendar, Notion, Stripe, Airtable or Clay through MCP app connectors, and you choose which tools each app may use. Voice AI agents can also connect to your own MCP servers through a Labs option, with conditions that set when each tool runs during a call. If you uninstall an app from the marketplace, it disconnects for every user in that sub-account.