Integrations
Built to plug right into your stack
Connect your whole stack in an afternoon. Read access by default, write access when you’re ready.
See if Cleric is a fit for your stack
Catch failures before your customers do
Cleric checks your changes in production and investigates regressions, bringing you fix PRs with the evidence to review them.
Read-only by default ·
SOC 2 Type II
Tighten timeout handling
Resolved#1849 @maya v2.14.0 Deployed Sep 15 at 09:41 # eng-checkout checkout-api
Summary
Strict timeouts for checkout's downstream calls
Expectations
Checkout calls use the new timeout Verified
Checkout p95 latency stays at baseline Anomaly found
p95 latency 182 ms → 511 ms
200 ms 400 ms PR #1849
09:2009:3009:4009:5010:0010:1010:2010:30
The fix
ResolvedPR #1893: Revert CHECKOUT_V2_STRICT_TIMEOUT
- 10:22 Cleric opened the fix PR
- 10:31
Approved by
Maya - 10:52 Deployed and verified, p95 back to 180 ms
Thread #eng-checkout
Cleric APP 9:50 AM
⚠️ Anomaly found: p95 up after PR #1849
1 reply
Cleric APP 10:52 AM
✅ Resolved: the regression I flagged has been fixed.
- Intent: strict timeouts, verified
- Regression: p95 +329 ms, now fixed, back to 180 ms
Catch what your tests miss
Cleric checks how your changes behave under real traffic and investigates the regressions your tests missed.
Track your PR into production
Cleric analyzes your PR, then tracks its deployment to check that production behaves as you intended.
PR #1849 · Verification record Expectation not met
Expected in production
p95 latency ≤ 200 ms, unchanged Retry rate ≤ 0.5%, unchanged
Observed
Deployed 09:41, checked 10:14
p95 latency 511 ms · expected ≤ 200 Retry rate 4.1% · expected ≤ 0.5
Catch delayed regressions
Cleric keeps checking your changes for days after deployment, letting you know when it finds a regression or verifies your release.
PR #1866 · Verification history Day 9 of 14
- Day 1 orders-api deployed · checks passed
- Day 3 Passed
- Day 9 Deviation · cache hit rate falling
Sent to the team
Warm-up covers a fixed key set. The working set has grown past it since deploy.
Cache hit rate 94% → 61% Orders p95 140 ms → 390 ms
Only the pages that need you
Cleric handles your urgent alerts and investigates low-priority issues before they interrupt your day.
Leave the pager to Cleric
Cleric acknowledges your pages, closes false positives, and resolves what it can, escalating to you only when you’re needed, with the findings attached.
P1 · payment-api 5xx Escalated
What Cleric did
- 02:14 Acknowledged the page
- 02:19 Ruled out deploy, config, and capacity
- 02:26 Confirmed retry storm from PR #470
- 02:31 Escalated with the investigation
What reached the engineer
Service payment-api Why a person Rollback needs a schema decision Decision needed Roll back, or cap retries and hold Evidence attached 7 signals · 3 ruled out
Find the issues buried in the noise
Cleric deduplicates your low-priority alerts and investigates the underlying problems, so you can get them resolved before they become urgent.
Low-priority stream · 6 days P3 · P4
23 alerts came in
auth-svc slow response 18 duplicates Session cache evictions 3 related
Became one issue
INV-1251auth-svc session lookups degrading
Cause Session TTL shortened in PR #1871 Fix PR #1907 opened
First alert Day 1, fix Day 6 No urgent page
Fix it your way
With the investigation done, you can finish the fix in Cleric or take the findings and production context to your coding agents.
Fix it in Cleric
You can review the fix PR and its production evidence in Cleric before approving the change. Once it deploys, Cleric checks that your issue is resolved.
INV-1247 · Fix review Resolved
The evidence you review it against
Cause Retry storm, PR #470 Confirmed by 7 signals · 3 causes ruled out
PR #483 payment-api
-max_retries: 5 +max_retries: 2 +backoff: exponential - 14:26 You approved it, in Cleric
- 14:31 Deployed · build 9c1d044
- 14:38 Recovery verified · 5xx back to baseline
Use your existing tools
Ask your coding agent to pull the investigation from Cleric over MCP, or have Cleric create a Linear, Jira, or GitHub issue with the findings and production context.
INV-1251 · Handed off Over MCP
What your coding agent pulled
Issue auth-svc session lookups degrading Finding Session TTL shortened in PR #1871 Evidence Eviction rate · p95 · TTL diff Existing fix PR #1907
claude-code > picked it up, starting from PR #1907
Or raise it as a Linear, Jira or GitHub issue. The investigation goes with it either way.
Prevent incidents in your own stack
Try Cleric on your production changes to see the regressions it finds in your stack and review the fixes it writes.