MCP Servers

MCP (Model Context Protocol) is an open standard for connecting AI agents to external tools. Astonish supports MCP servers that provide additional capabilities beyond the built-in tool set.

Configuration

MCP servers are configured in ~/.config/astonish/mcp_config.json. Two transport modes are supported: stdio (local process) and sse (remote HTTP).

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_TOKEN": "ghp_..." },
      "enabled": true
    },
    "remote-server": {
      "transport": "sse",
      "url": "https://example.com/mcp",
      "enabled": true
    }
  }
}

Installing from the MCP Store

Studio includes a built-in MCP store browser. From the CLI, use:

astonish tools store

Managing Servers

Command	Description
astonish tools servers	List all MCP servers with status
astonish tools enable <name>	Enable a server
astonish tools disable <name>	Disable a server
astonish tools refresh	Reconnect and refresh tool cache
astonish tools edit	Edit MCP config directly

How Tools Appear

Tools from MCP servers appear alongside built-in tools. The agent uses them seamlessly without any distinction — you simply describe what you need, and the agent selects the appropriate tool whether it is built-in or provided by an MCP server.

Flow Dependencies

Flows can declare MCP dependencies that get auto-installed when the flow runs. See the YAML Reference for the mcp_dependencies syntax.