Skip to content

Repository files navigation

Handoff Workbench

Handoff Workbench makes authority transfer between AI agents explicit, reviewable, and verifiable. A human sees the exact proposed scope, approves its digest, and can follow the handoff through a versioned lifecycle to a verified completion receipt.

Open the live synthetic Workbench

Handoff Workbench

What it demonstrates

  • Signed handoff envelopes with explicit recipient, capability, resource, data, money, evidence, and expiration bounds
  • Exact digest-bound human approval before delegation advances
  • A seven-step lifecycle with seven ordered durable events at completion
  • Receipt verification and durable recovery after reload
  • Stable rejection codes with no state change for demonstrated authority expansion, wrong-recipient, stale-version, replay, and tampering attacks
  • A narrow MCP tool boundary for programmatic integration

The supplied scenario is synthetic. It does not represent customer adoption, production financial authority, unsupervised operation, or live-network rate exhaustion.

Run locally

Requirements: Node.js 24.12.x and npm 11.6.x.

npm ci
npm test
npm run workbench

Open http://127.0.0.1:4188/. The documented command starts the interface and its real local application service on one same-origin listener. SQLite state is stored under .workbench-local/ unless WORKBENCH_DB_PATH is set.

Explore

Agent Handoff API is a Wever Labs project. Copyright 2026 Wever Labs. No open-source license is granted.

About

Verifiable authority handoffs for AI agents

Resources

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages