Skip to main content

Sentry OpenCode Debugging

Triage a production error from Sentry using the hosted MCP
(https://mcp.sentry.dev/mcp via OAuth) and The Debugger's five-step protocol.
Portal setup: docs/portals/sentry.md.

Steps

  1. Verify access — ask What organizations do I have access to in Sentry?

    • If auth fails, re-run opencode mcp auth sentry and retry once

  2. Search by impact — search_issues filtered by time range, sorted by impact

    • Prefer issues affecting the most users first
    • Example: issues from the last week affecting more than 10 users

  3. Read details — get_issue_details for the full stack trace, events, release tag

    • Correlate the release tag with git log for that tag
    • State the exact error, file, and line before moving on

  4. Targeted fix — apply the five-step protocol (reproduce → hypothesize → test → fix at root cause)

  5. Systemic? escalate to Seer — check the escalation criteria in
    docs/portals/sentry.md ("Seer escalation criteria"); if met, run
    begin_seer_issue_fix and record the plan link in the PR description

  6. Regression test — write a test that fails on unfixed code and passes on fixed code;
    document the root cause in the PR description

Notes

  • Scoped to issue/PR reads and comment writes (issues: write covers issue comments in GHA); sibling templates still use read-all (migrate them separately if the Society adopts least-privilege frontmatter)
  • Does not modify the src/portals/ runtime — MCP tools only
  • If the hosted endpoint is unreachable, fall back to the legacy npx stdio config in docs/portals/sentry.md
  • Flags security-relevant errors (auth failures, injection attempts) for The Auditor