New York · October 12–14
A decision surface shared with ChatGPT.
Inspect the plan, repair its constraint, and keep consequential approval in human hands.
- Budget
- $7,200
- People
- 8
- Late arrival
- 11:40
Active decision
Can the roadmap session include everyone?
- Inspect
- Repair
- Review
- Reserve
ChatGPT and this page work on the same live plan. You authorize the exact version before anything is reserved.
Reservation confirmed
The exact authorized plan is reserved.
Restoring this receipt never creates a duplicate reservation.
Start with the first proof. Ask ChatGPT to find what blocks this plan.
Find the most important conflict in this offsite plan.- Roadmap decisionsAll 8 people required
- Working lunchAll 8 people
- Operating model6 people required
Conflict found
Jump to conflictFeasible repairs
Two ways to include everyone
| Measure | Shift session | Remote access |
|---|
Human authorization
You are approving exactly this version.
What changes?
What does it cost?
What are you authorizing?One simulated offsite reservation
ChatGPT can prepare this review but cannot authorize it.
Technical details
- Plan
- Scope
- reserve_offsite
- Expiry
- Created after authorization
Action not completed
Technical evidence
Protocol run
Collecting- Page tools
- Awaiting WebMCP
- Tool version
- Awaiting registration
- Agent saw current tools
- Not reported
- Old tool calls blocked
- Not tested
- Unauthorized executions
- Duplicate reservations
- Not tested
- Visible update time
- No samples
- Receipt restored
- Not observed
Each tool set is bound to its issuing revision. Obsolete callbacks are rejected even if a host retains them. Discovery and removal acknowledgement are not exposed by the current API, so agent observation is explicitly reported.
Raw protocol trace
- No protocol events yet.
Decision trail
- No structured actions yet.