AgentApprovals

Know when an
agent should ask.

What may it do? When should a person decide? How do you know what actually happened?

Independent guides for developers building AI agents.

Start with the decision guide

The guides

Practical decisions. Sources you can inspect.

ImplementationPublished

Did your agent commit the change you approved?

Capture the staged Git tree and its base before review, then check the finished commit against both. Includes a tested shell verifier and an agent handoff.

Includes: Git commit review commands and verifier

↗
WorkflowPublished

You stopped the agent. Did its action stop?

Separate cancellation requests from stopped execution and verified effects. Use a worked batch example and an editable outcome worksheet before retrying.

Includes: Cancellation outcome worksheet

↗
WorkflowPublished

Does a read-only MCP tool still need approval?

Evaluate MCP tool annotations before granting automatic approval. Follow a ticket-search example and use an editable tool review worksheet.

Includes: MCP tool approval review worksheet

↗
WorkflowPublished

Write an AI agent approval request with a clear scope

A copyable approval request template for AI agents. Show the exact change, destination, expiry, and checks, then keep the decision separate from the result.

Includes: Approval request card and event reference

↗

Read the reasoning.
Check the evidence.

We separate documented behavior from proposed designs. Every guide should leave you with something useful: a decision table, a reproducible example, or a clearly labeled experiment.

How we research and review →