DEVELOPMENT

Sentry errors resolved before your team sees them

Agents monitor Sentry for new errors, investigate root causes in the codebase, propose fixes, create pull requests, and even deploy patches — all automatically. Wake up to fixed bugs, not new tickets.

What you get

Automated error monitoring

Sentry issues detected and prioritized automatically. The agent checks for new errors, ranks them by frequency and user impact, and focuses on what matters most.

Root cause investigation

The agent reads stacktraces, finds the relevant source code, traces the bug through the codebase, and identifies the root cause — not just the symptom.

Automated fixes with PRs

Code changes proposed with clear explanations. Each pull request includes a description of the bug, the root cause, and why the fix works.

Human-in-the-loop for complex bugs

The agent flags what it can't confidently fix — providing investigation notes, likely root cause, and recommended next steps for your team to review.

Continuous improvement

Recurring daily runs catch regressions immediately. Bugs introduced yesterday are investigated and fixed today — before they compound into bigger problems.

Done by an agent. Zero human effort.

Schedule it daily. The agent checks Sentry overnight, investigates errors, creates PRs for fixable bugs, and drops a summary in Slack before your team's morning standup.

Developer Agent · Bug Triage Complete
> Checking Sentry for new errors since last run
> Found 12 new issues — prioritizing by frequency & impact

Investigating top 5 errors:
> #1 TypeError in /api/payments — reading stacktrace, tracing source
> #2 ConnectionError in /api/sync — identifying retry logic gap
> #3 ValidationError in /api/users — null check missing on input
> #4 RangeError in /lib/pagination — off-by-one in cursor calc
> #5 TimeoutError in /api/export — needs architectural review

Results:
> Fixed 3 of 5 errors — PRs created with explanations
> Flagged 2 errors for human review — investigation notes attached

✓ PR #412: Fix null pointer in payment processing
✓ PR #413: Add retry logic for sync connection errors
✓ PR #414: Add input validation for user endpoint
✓ Summary posted to #engineering on Slack
Duration: 32m · Next scheduled run: tomorrow 2:00 AM
HOW TO SET IT UP

For CRHQ clients

1

Enable the required skills

Make sure these skills are active on your satellite:

2

Connect Sentry and define your workflow

Example prompt:

"Check Sentry for new errors in our main project. Prioritize by frequency and user impact. For the top 5 errors, investigate the root cause by reading the stacktrace and finding the relevant source code. If you can confidently fix the bug, create a pull request with a clear description. If you can't fix it, write up investigation notes and flag it for human review. Post a summary of everything to #engineering on Slack."
3

Schedule daily runs

Set up a recurring daily schedule so the agent triages bugs overnight. New errors are investigated and fixed before your team starts work each morning. Use the schedule skill to automate it.

ADVANCED TIPS

Get more out of this workflow

Pair with automated code review

Once the agent creates a PR, have another agent (or the same one) run automated code review to check the fix for quality, test coverage, and potential side effects. This creates a double-check before any human reviews the change.

Enforce project standards with coding-best-practices

The coding-best-practices skill ensures every fix follows your project's coding standards — naming conventions, error handling patterns, test requirements, and documentation. Fixes that don't meet the bar are flagged before the PR is created.

coding-best-practices

Delegate complex investigations to sub-agents

For bugs that span multiple services or require deep codebase exploration, use the delegation skill to spawn specialized sub-agents. One agent traces the frontend, another traces the API, and a third checks the database layer — all coordinated by a lead agent.

delegation

Build a knowledge base with investigation notes

Store investigation notes in project documents so future triage runs can reference past findings. Over time, the agent builds institutional knowledge about your codebase — common failure patterns, fragile areas, and recurring issues.

Ship with confidence.
$299/month per satellite. Bring your own AI model. Deploy in 90 seconds.