Connectors (MCP)
Plug external MCP tool servers into the platform so agents can use third-party tools.
Connectors let the platform talk to external tool servers via the Model Context Protocol (MCP). Each connector is an MCP server you register; every tool that server exposes automatically appears in the platform's tool catalog under a group named after the server, ready to be granted to any agent.
Common examples: filesystem access, GitHub/Jira integration, database inspection, browser automation - anything with an MCP implementation.
Transport types
Connects to a remote MCP server over HTTP Server-Sent Events. Use this for hosted/shared tool services or servers running elsewhere on your network - provide the server's endpoint URL.
Adding a connector
Open the Connectors section
Navigate to Settings → AI → Connectors and add a new MCP server config.
Describe the server
Give the connector a name and short description. These appear in the tool catalog group and help agents (and humans) understand what the tools are for.
Configure the transport
- Pick
sseand enter the endpoint URL, or - Pick
stdioand enter the command plus its arguments JSON.
If the server requires credentials, store them as encrypted environment variables or headers as prompted by the form.
Activate and select tools
Mark the connector active. Its tools are discovered automatically and listed under the connector's group in the tool catalog. As with built-in tools, you control per-agent access via the tool selector on each agent.
Trust external tools
An MCP server's tools execute with the capabilities that server provides. Only connect servers you trust, and grant their tools only to agents that genuinely need them.