Connect its tools
Tell safe reads from consequential writes
Apply approval and governance rules from an explicit effect classification.
Classify each tool so approval policy can target the actions that may change data.
What this changes for your team.
Each tool declares whether it reads or writes external state. The platform uses that declaration to expose risk clearly and to decide which calls are eligible for write-tool approval gates; the declaration cannot prove what a misdescribed third-party endpoint actually does.
How it works in practice.
- 01
The integration author classifies each operation when defining the tool.
- 02
Agent authors see the classification while choosing tools and approval policy.
- 03
The runtime applies write-specific approval behavior before dispatching a classified write call.
What you can plan around.
The behaviour you can design against, stated concretely.
Effect behavior is integration metadata evaluated by server-side execution code.
Read tools are never silently promoted to writes, and ungated declared writes remain visibly ungated.
For custom integrations, the organization that authors the definition is responsible for an accurate classification.
Bring one real process
See how Yekar.AI fits the way you work.
Start with a job your team already owns, plus the tools and decisions around it.
Talk to us