Connect to Model Context Protocol servers for rich tool integrations. MCP servers provide structured access to services like GitHub, Linear, and Notion.

What is MCP?

The Model Context Protocol (MCP) is a standard for connecting AI models to external tools and data sources. MCP servers expose tools that TorlyAI can use during conversations, enabling actions like creating GitHub issues, querying databases, or managing tasks in Linear.

Popular MCP Servers

GitHub

Create issues, PRs, manage repositories

Linear

Create and manage tasks and projects

Notion

Access and update Notion databases

Slack

Send messages and manage channels

Filesystem

Read and write local files

PostgreSQL

Query and update databases

Adding an MCP Server

Click Sources → Add Source
Select MCP Server
Choose the transport type (stdio, HTTP, or SSE)
Enter the server command or URL
Configure authentication if required
Test the connection

Transport Types

stdio

Runs a local process and communicates via stdin/stdout. Best for local MCP servers.

{
  "transport": "stdio",
  "command": "npx",
  "args": ["-y", "@modelcontextprotocol/server-github"]
}

HTTP

Connects to a remote HTTP endpoint. Good for hosted MCP servers.

{
  "transport": "http",
  "url": "https://mcp.example.com/api"
}

Configuration File

MCP server configurations are stored in:

~/.torlyai/workspaces/{id}/sources/{slug}/config.json