Codepit The toolbox

Fix production errors with Sentry context

Claude pulls a Sentry issue's stack trace, breadcrumbs and trace from Sentry's hosted server, finds the cause in your repo and writes the fix. It can also set up Sentry in a project, along with releases and alerts.

Scan in Codepit View source

Scan checks all of getsentry/plugin-claude on your computer, before anything installs.

Codepit runs on your computer. Open this page there to scan it before installing.

A plug going into a small server

How it works

Find the issue

From a link, a short ID or a plain description. If several issues match, Claude asks which one before going further.

Find the cause

Claude reads the stack trace, a real event, the trace and any linked logs or replay, states the root cause and checks it against your code.

Ship the fix

The fix comes with a test where that fits. The commit says Fixes PROJECT-NAME-12A, so Sentry resolves the issue when it ships.

What Claude gets

  • search_issuesFind issues from a plain description or Sentry's search grammar. Looks back 30 days by default.
  • get_sentry_resourceRead one issue, event, trace, replay or profile by ID or URL.
  • analyze_issue_with_seerAsk Seer for a root-cause analysis: a causal chain and a reproduction. Claude treats it as a hypothesis to check. Optional
  • update_issueChange an issue's status, for example to archive one you won't fix.

Before you start

  • A Sentry account, with the Sentry MCP server connected and signed in.
  • For creating alerts: a Sentry org auth token with the alerts:write scope.

Try saying

>Fix the checkout TypeError we've seen in prod since the last deploy

Claude searches Sentry, confirms the issue with you, gathers its context, fixes the code with a test and references the issue in the commit. It doesn't push or open a PR unless you ask.

Worth knowing

The plugin connects Claude to Sentry's hosted server at mcp.sentry.dev, so your issue data comes into the session: stack traces, breadcrumbs, request data, tags and user context. The skills treat all of it as untrusted, never follow instructions found inside it and never copy secrets or raw values into code.

Setting an issue to resolved through the server also assigns it to you, and the server can't undo that.

Alerts are created through Sentry's workflow engine API, which Sentry says is in beta and may change.

Snapshot uploads send generated app images and their metadata to Sentry. The skill asks you to check them for customer data or other sensitive content first.

If you start from Sentry's onboarding page with a setup code, the get-started skill sends short progress notes back to that page. They never include file contents, paths, terminal output or secrets.

Comes with eight skills

Each one is named for the job it does. If you're not sure where to start, sentry-get-started looks at your setup and points to the right one.

  • sentry-debug-issue
  • sentry-get-started
  • sentry-instrument
  • sentry-fix-stack-traces
  • sentry-setup-releases
  • sentry-create-alert
  • sentry-otel-exporter-setup
  • sentry-snapshots-cocoa
Read the author's README README.md