Control
Limit an AI agent to the exact systems, targets, and supported actions you authorize.
What TookEffect is
TookEffect verifies what AI agents actually changed.
TookEffect is an independent verification and control layer for AI agents. When an agent says it merged code, deployed software, or changed infrastructure, TookEffect checks the real external system to confirm the expected outcome actually happened and keeps evidence of the result.
The problem
An AI success message is not proof that the real-world outcome happened.
An API can accept a request, an automation can return success, and an AI agent can report “done” while the final system state is delayed, wrong, incomplete, or uncertain. TookEffect separates the claim from the outcome by reading the real system again before a workflow treats the action as verified.
How TookEffect works
Request → act → independently check → prove.
For supported actions, TookEffect identifies the exact target and expected effect, applies the authorized action, independently observes the external system, decides whether the expected result took effect, and stores evidence plus a verifiable receipt.
Verify
Check the real provider state instead of trusting the agent or the first API success response.
Prove
Keep the observed result, decision, evidence, and signed receipt for later inspection.
Available verification paths
One verification model across external systems.
TookEffect currently supports verified actions across GitHub, Vercel, and Cloudflare Workers. The verification layer is designed so additional providers and AI agents can use the same outcome-checking model as new paths are certified.
AI agent merge verification
Verify that an exact pull request really merged into the expected branch.
Deployment promotion verification
Verify that production aliases point to the deployment the workflow intended to promote.
Deployment verification
Verify the active Worker deployment and resulting traffic state after an authorized change.
Why teams use it
AI agents can act. TookEffect gives the action an independent check.
TookEffect is infrastructure for consequential AI-agent actions: action control, independent outcome verification, evidence, and receipts. It is not another AI model and it does not require trusting the agent's own report of success.
Verification model
CHECK REALITY- Agent claim
- “The action is done”
- TookEffect check
- Read the external system again
- Decision
- Applied, not applied, or uncertain
- Evidence
- Observed state and reason
- Receipt
- Verifiable record of the result
Independent verification for AI agents
Do not stop at “done.” Verify what took effect.
Start with one supported action and inspect the evidence TookEffect keeps after checking the real system.