A knowledge base built from documents and URLs can answer most support questions. But some queries require live data: what is the current status of this order, what is the latest price for this product, or has this issue been logged in our ticket system? That's where MCP servers come in.
What is the Model Context Protocol?
MCP (Model Context Protocol) is an open standard that defines how AI models communicate with external data sources and tools. An MCP server exposes a set of "tools" — functions the AI agent can invoke at runtime. The agent discovers those tools automatically when a conversation starts, and calls them when relevant.
What can you connect via MCP?
- Notion — let the assistant read and reference your Notion knowledge base in real time.
- GitHub — query issues, pull requests, or repository status directly from the conversation.
- Internal databases — expose read-only query tools via a custom MCP server.
- Custom APIs — wrap any internal service as an MCP tool without changing existing infrastructure.
- Project management tools — Jira, Linear, Asana, and others via community MCP servers.
Authentication options
MCP servers often require authentication. The platform supports Bearer token, Basic auth, OAuth 2.0 Client Credentials, and OAuth 2.0 Authorization Code with PKCE. For PKCE flows, the OAuth endpoints are auto-discovered from the server URL — you don't need to manually configure authorization or token endpoints in most cases.
// Example: testing your MCP server before going live // The platform sends a discovery request to your server URL // and returns the list of tools it found // On success: // ✓ Connected — 12 tools found // On failure: // ✗ Connection failed: unable to reach server at https://...
MCP server connections are a Pro feature. You can connect multiple servers simultaneously — the agent queries all enabled servers in parallel and decides which tools to invoke based on the user's intent.