MCP Server

TrapFall includes a Model Context Protocol (MCP) server for AI agent integration. Communicates via stdio JSON-RPC 2.0.

Starting

trapfall mcp --db trapfall.db

Configure your AI client to run this command as an MCP server.

Available Tools

Tool	Description
list_issues	List issues for a project (with filters)
get_issue	Get issue detail by ID
get_event	Get event detail by ID
set_status	Set issue status (resolved/unresolved/ignored)
search_issues	Search issues by query string
list_projects	List all projects
get_project	Get project detail by slug
get_project_stats	Get issue/event counts for a project
list_alert_rules	List alert rules for a project
list_events	List events for an issue
rotate_dsn	Generate new DSN key for a project
healthcheck	Check server health

Example: AI Agent Usage

An AI agent connected via MCP can:

1. Monitor errors — list_issues to check for new errors
2. Investigate — get_issue + list_events to see full stack traces
3. Take action — set_status to resolve or ignore issues
4. Search — search_issues to find specific errors
5. Manage projects — list_projects, get_project_stats

Transport

stdio only. No TCP. This is by design — MCP servers are launched as child processes by the AI client.