Skip to main content
SafeFetch includes a built-in Model Context Protocol (MCP) server. This lets AI assistants send actions, check status, and manage approvals directly through their tool interface.

Setup

1. Get your account ID

Your account ID is shown on your dashboard. It looks like acct_abc123.

2. Configure your MCP client

Add SafeFetch to your MCP client config. The server runs over stdio. Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):
Cursor (.cursor/mcp.json in your project root):

3. Restart your client

The SafeFetch tools will appear in your assistant’s tool list.

Available tools

send_action

Send a new action. Returns immediately with an action ID, or set sync: true to wait for completion.
Parameters:

get_action

Get the current status and result of an action.

list_actions

List recent actions, optionally filtered by status.

approve

Approve an action waiting for human sign-off. Moves it to pending for immediate dispatch.

cancel

Cancel a pending or awaiting-approval action.

retry

Retry a failed or cancelled action. Creates a new action with the same parameters.

Example conversation

Action statuses