Skip to main content

MCP Servers

Model Context Protocol (MCP) servers are another way to connect Abacus.AI to external tools and data sources. MCP is an open standard that lets AI agents and chatbots talk to external systems — APIs, databases, and services — through a common protocol, without building a custom integration for each one.

MCP servers complement the First Party Connectors and the other connector types described in this section. Use an MCP server when the tool or service you need is exposed as an MCP server (community-built or custom) and is not already available as a First Party Connector.

Two ways to connect​

  • User-level MCP servers — any member can add an MCP server for their own use and authenticate with their own credentials. This is the quickest way to try a single server.
  • Organization-level MCP servers — an administrator configures a server once for the whole organization. The configuration is shared with every member, but no credentials are shared: each member connects and authenticates with their own account. This gives you standardized, central setup across the team.

Both types share the same limits: up to 5 active servers and up to 50 tools across those servers. Servers that need access to your local filesystem will not work, because MCP servers run in an isolated environment.

Where to configure​

MCP servers are configured in ChatLLM Teams, under Settings → Connectors → MCP Server Configuration. From there you can add a server with a guided form or JSON, and — if you are an admin — share it with your whole organization.

Using MCP servers with Custom Chatbots​

Once you have connected to an MCP server, its tools can be attached to a Custom Chatbot so the bot can act on that external service. Note that a chatbot using MCP servers (or Connector Tools) cannot be made Public, because it relies on per-user credentials.

Learn more​