Skip to main content

Bugfender MCP

Bugfender MCP exposes Bugfender data through the Model Context Protocol so AI agents can investigate logs, network requests, crashes, issues, devices, feedback, and app metadata directly from your IDE or CLI.

The recommended connection is Bugfender's hosted MCP service:

https://mcp.bugfender.com/mcp

Hosted connections use Bugfender OAuth. You sign in on the Bugfender dashboard, review the requested permissions, and approve the connection without copying an API token into the client.

What it provides​

  • User-scoped Bugfender access
  • OAuth sign-in and automatic token refresh for hosted connections
  • Tools for logs, devices, apps, crashes, issues, feedback, and SDK snippets
  • Network aggregate and request-detail tools
  • An explicit tool for changing an issue group's status
  • Resource templates for snippets, logs, and issue groups
  • Companion skills for investigation and SDK setup

Choose a connection method​

Bugfender Claude Plugin​

Install the Bugfender Claude Plugin to add the hosted connection together with the investigation and SDK-setup skills. See the dedicated Claude Plugin and Connector guide for supported products, installation, OAuth, and troubleshooting.

Bugfender Codex Plugin​

Install the Bugfender plugin in ChatGPT or Codex, select Connect, and complete the Bugfender consent screen. It includes the same hosted MCP connection and companion skills, packaged for Codex.

Direct remote MCP​

Clients that support remote Streamable HTTP and OAuth can connect directly to:

https://mcp.bugfender.com/mcp

The client discovers Bugfender's OAuth endpoints automatically. Do not paste a Bugfender token into a hosted connection.

Legacy local stdio​

The @bugfender/mcp package is deprecated for new installations. Existing local and self-hosted integrations can continue running it with:

npx -y @bugfender/mcp

See Client Setup for the retained environment-variable configuration. Hosted OAuth tokens and local stdio tokens are separate and cannot be exchanged between the two modes.

OAuth scopes​

  • mcp:read allows identity and team lookup plus read-only access to apps, SDK metadata, logs, devices, crashes, issues, feedback, network data, and app summaries that the signed-in user can already access.
  • mcp:issues:write allows only the update_issue_status tool. It does not grant general app, team, account, or billing administration.

A read-only connection can request only mcp:read. The issue-status tool then remains visible but requires reconnecting with the write scope before it can run.

Tokens, disconnection, and revocation​

Hosted access and refresh tokens are bound to the signed-in user, OAuth client, Bugfender MCP resource, and granted scopes. They cannot be reused for another client or audience, and refresh cannot increase permissions.

Disconnecting Bugfender revokes the hosted grant and its tokens. A later tool call must start a new OAuth connection. Removing the plugin without disconnecting may only remove the local installation, so disconnect the Bugfender connection first when the client presents those as separate actions.

Privacy and data processing​

The hosted MCP forwards requests to the Bugfender API and returns only data the signed-in user is authorized to access. Access tokens are request-scoped and are not written to the hosted MCP filesystem or included in tool responses. Your MCP client and AI provider may process tool inputs and return Bugfender data under their own privacy and retention terms.

Review the Bugfender Privacy Policy and Terms of Service before connecting.

Next steps​