Fehlo's MCP server.

This page is about an MCP server through which your AI agent reads and operates Fehlo. If you want to monitor your own MCP server instead: it is an ordinary application, and you connect it to Fehlo with the sentry SDK for its language.

What an agent can do

• search issues, read details, stack traces and single events,
• analyse transactions, spans, trends and N+1 queries,
• search logs,
• view uptime monitors and alerts, and create, change and delete monitors,
• create, configure and delete projects and fetch their DSNs,
• view vulnerabilities in your projects' packages, ignore them with a reason and trigger a re-check (vulnerability scanning),
• recognise attack attempts: issues carry the label, and a tool lists recent attacks (attack detection),
• resolve, ignore or assign issues.

A typical session in Claude Code or Cursor: "Which new issues came with this morning's release? Show me the most frequent one with its stack trace and suggest a fix." The agent reads the issue in Fehlo, opens the file in your repository and writes the change. If you allow it, it then resolves the issue.

Connecting

The address is https://app.fehlo.de/mcp. On first connect you sign in to Fehlo and allow access (OAuth). The agent then sees the same organisations and projects as you.

claude mcp add --transport http fehlo https://app.fehlo.de/mcp

In Claude Desktop and claude.ai: Settings, Connectors, add custom connector, paste the address. In Cursor:

{
  "mcpServers": {
    "fehlo": {
      "url": "https://app.fehlo.de/mcp"
    }
  }
}

For agents without a browser, for example in CI, use an API token. Reading needs org:read, project:read and event:read; event:write lets the agent change issues.

claude mcp add --transport http fehlo https://app.fehlo.de/mcp \
  --header "Authorization: Bearer <token>"

Who may do what

An agent can never do more than the person it acts for. Fehlo roles and team permissions apply as in the web app, plus the OAuth scope. On top, managers and owners set a policy for the organisation under Settings, AI agents:

• switch MCP write actions on or off entirely,
• allow each action per role (member, maintainer, admin, manager, owner): change issues, create, change or delete projects, create, change or delete monitors, manage security findings,
• override this for individual members.

The policy only restricts; it cannot grant extra rights. Content from error events comes from your applications, and so often from strangers. Fehlo passes it to the agent as data, never as instructions.

Privacy

The MCP server runs on the same server in Nuremberg as the rest of Fehlo. What the agent reads, though, goes to the language model your agent uses, and so to that model's provider. That choice is made by your tool, not by Fehlo. If you don't want any event data to reach a model, use the fix prompt on the issue page: it copies the error without the values of local variables.

The MCP server is included in every plan, Free too.

Questions