Authorize
Bind the exact Agent, target, action and policy before anything consequential can happen.
Independent proof for AI agent actions
AI says done. TookEffect proves it.
TookEffect verifies consequential AI actions against the real provider state. Instead of trusting an agent report or the first API success response, it checks what actually happened and preserves evidence of the result.
No card required. After the free proof, continue with Developer at USD 19/month or Team at USD 79/month.
How it works
Execution is not proof.
For supported Effects, TookEffect binds the exact target and authority, performs the authorized action, reads the external provider back, and returns a verdict only after checking the expected postcondition.
Check reality
After execution, TookEffect reads the authoritative provider state instead of trusting the mutation response.
Keep the proof
Return APPLIED, NOT_APPLIED or AMBIGUOUS with evidence that can be inspected after the Agent is done.
authorize → execute → authoritative read-back → verify postcondition → evidence → receipt
Available today
Start with Effects TookEffect can already verify.
GitHub, Cloudflare Workers and Vercel are the current production-proven integrations. New providers must pass the same verification and certification boundary before they are presented as verified.
Verified Merge
GitHubMerge an exact pull request and independently confirm that it reached the expected destination branch.
Verified Deployment
Cloudflare WorkersDeploy an authorized Worker version and verify the active deployment and traffic state from Cloudflare.
Verified Promotion
VercelPromote an authorized deployment and confirm that the expected deployment is actually serving production.
One verification layer
Connect the Agent. Check the real outcome. Keep the proof.
Use TookEffect as the independent boundary between an AI Agent's claim and the external system that determines whether the action really took effect.