Skip to main content
Connect Model Context Protocol (MCP) servers to extend your agents with external tools and data sources.
Prerequisites: API key, MCP server endpoint URL

Overview

MCP servers provide a standardized way for agents to access external tools and data sources. Agents can connect to multiple MCP servers to extend their capabilities beyond built-in integrations.

Configuration Fields

  • name (string, required) - Human-readable name for the server
  • description (string, optional) - Description of the server’s purpose or tools
  • url (string, required) - MCP server endpoint URL
  • auth_type (string, optional) - Authentication type: "none", "bearer_token", "api_key", "custom_headers". Default: "none"
  • auth_token (string, optional) - Token for "bearer_token" or "api_key" authentication
  • headers (object, optional) - HTTP headers for authentication or custom configuration

Transport Type Detection

The API automatically detects the appropriate transport protocol based on the URL path:
  • URLs ending with /mcp → streamable HTTP transport (recommended)
  • URLs ending with /sse → Server-Sent Events transport (deprecated)
  • Other URLs → defaults to SSE for backward compatibility

Create Agent with MCP Servers

Configure MCP servers when creating a new agent.

Update Agent with MCP Servers

Add or modify MCP servers for an existing agent. Note: Agent must be paused before updating.

Authentication Types

No Authentication

Bearer Token

API Key

Custom Headers

Multiple MCP Servers

Agents can connect to multiple MCP servers simultaneously.

Best Practices

  • URL Format: Use URLs ending with /mcp for recommended HTTP transport
  • Security: Store authentication tokens securely and never expose them in client-side code
  • Error Handling: MCP servers that fail to connect are logged but don’t prevent agent deployment
  • Testing: Test MCP server connectivity before deploying agents to production

Next Steps