AI-native reference application
Ask naturally.
Execute deliberately.
A production-shaped spatial agent where every Mappls tool call is planned, risk-labelled, explicitly approved, scope-bound, and cited.
Human approval is a state transitionNo plan can execute until its exact hash and every requested scope are approved.
Run inspector
No run yet
AI
Governance before automation
Create a plan to see tool dependencies, risk, scopes, immutable hash, and the evidence produced by each Mappls operation.
02Review the execution plan
03Approve access
Explicit, expiring grant✓
Plan approved
04Grounded answer
Provider evidence attachedDeterministic, inspectable steps with an immutable SHA-256 identity.
Operators approve every scope required by the reviewed plan.
Answers cite the exact Mappls operation, region, host, and retrieval time.
Idempotent commands, version checks, events, and outbox survive restart.