Validation for agents
Clanker Test makes it fast to validate the work your agent just did. Write tests, collect evidence, and review the run in one place.
Agents write the tests. You review the proof.
Test run
Checkout flow
Passed · 12s · 4 steps
Built for a different author
Manual QA was built for people to click through. Clanker Test is built for agents to write validation tests, and people to review the evidence.
A local test run lives in a terminal. A screenshot lives in chat. Bring the cases, the evidence, and the review into one workspace.
See how it worksComing next
Today you can sign in, create a workspace, and invite your team. Tests, evidence, and scheduled runs come next.
A shorter path to proof
Ask it to validate the work it just did. The agent writes tests with the help of a plugin that teaches it what good validation looks like.
Prove checkout still works: sign up, add an item, and pay. Capture the evidence.
Open the run. See which steps passed, what the agent captured, and where it failed.
Payment succeeded, but the confirmation email never arrived.
The agent revises the tests from your feedback. Revisit earlier runs when you need to compare what changed.
Passed
Latest run
Email assertion added
Agent revision
Initial checkout coverage
Agent revision
Illustrative agent conversation, review, and test history.
Designed to be reviewed
A clear test case, evidence that answers the question, and a place to work through the failures.
Teach your coding agent how to write validation that a human can trust.
Provision inboxes, phone numbers, and secrets so agents can automate real flows.
Open a run and see the proof that the agent built what you asked for.
Invite the team, review results together, and keep the useful runs.
Pricing
Free
Solo
Teams
Scale
A few useful details
Your coding agent writes the tests. Clanker Test teaches it how, gives it the resources it needs, and gives you a place to review the results. Ask your agent to validate the work it just did.
Create an account and a workspace. A plugin and agent API come next. For now you can sign in, create workspaces, and invite people.
Local runs live in a terminal on one machine. Clanker Test is a shared place to publish results, inspect evidence, and show that the agent built what you asked for.
Later: disposable email inboxes, phone numbers for OTP, secrets, and other resources so agents can automate real product flows. Those APIs are not in this first cut.
Yes. Scheduled and automated runs come after the first product cut. This version is the logged-in workspace foundation.
Make the next change one you can verify.