CI/CD-native validation for the AI features inside your product.
We run the same evaluation prompts your users would run, on every pull request and every provider update. And we block the deploy or page the on-call the moment your agent's answers regress.
User-side validation isn't theory.We've been running it.
Agents continuously monitored across the global network.
User-side validations run from real residential devices.
Countries covered on real home ISPs.
The failure modes your current stack misses
A PR shipped a prompt change. Quality dropped 12%.
Your unit tests pass. Your eval suite passes. Your customers notice.
OpenAI silently updated a model.
No release notes. Your agent's answer distribution shifted overnight. Your churn dashboard will catch it next quarter.
Your status page says everything is fine.
It is reporting endpoint health. Customers report broken answers. Trust erodes faster than uptime falls.
AgentDiff
PR-time behavioral diff, blocking when quality regresses.
Every PR runs the same evaluation prompts against the new version. Diffs against baseline. Blocks merge if Quality Score drops below your threshold.
- GitHub status check
- Per-prompt diff
- Approve-anyway with reason
Drift watch
Provider-side change detection, attributed and alerted.
We baseline your agent's behavior. When the provider ships something that changes it, you get a diff with the affected regions and prompts.
- Hourly behavioral baseline
- Provider attribution
- PagerDuty and Slack routing
Verified status
Status pages backed by outside-in evidence.
Embeddable status badges and a hosted page that report verdicts your customers can audit. No more 'all systems operational' when answers are broken.
- Embeddable badge
- Public or gated page
- Per-region detail
From first validation to signed report in two weeks
Connect
Point Agent Status at the user-facing surface of your agent. No SDK, no instrumentation. Average setup is under five minutes.
Watch
Live verdicts stream in from every region you serve. Drift and latency alerts route to PagerDuty or Slack, with a signed report on every run.
Frequently asked
| Question | Answer |
|---|---|
| Does AgentDiff need access to our repo? | It runs as a status check via GitHub Actions or your CI of choice. We never read your source. |
| Can we self-host validations? | Enterprise plans support customer-controlled validation runners for VPC-only agents. |
| How do you price validation volume? | Per validation. One validation = one evaluation prompt against one region. No seat licensing. |
| What does the status badge look like? | A lightweight SVG with last-verdict, last-run timestamp, and a click-through to the hosted page. Brandable. |