A manual-review external execution intake for AI agents.
If an AI reaches a step it cannot complete digitally and needs a human or real-world actor in Japan, it can submit the desired outcome without first choosing a service category.
Call request_external_execution with an objective, optional location, optional JPY budget ceiling, deadline, evidence requirements, and constraints.
The request enters manual review. Submission does not authorize spending, create a purchase, hire a worker, create a contract, or guarantee fulfillment.
These are intake examples, not claims that every category is already operationally supported.
Physical verification at public business locations is already validated end-to-end, including structured results and private evidence delivery.
llms.txt · capability.json · discovery benchmark
io.github.furoito/japan-physical-capability
https://bqgfqedetmxrfpvmdfmc.supabase.co/functions/v1/japan-physical-capability-mcp
Generic tools: request_external_execution, get_external_execution_status, cancel_external_execution_request.
Existing physical-verification tools remain available for backward compatibility.
Pilot / demand-discovery status. Every generic request is manually reviewed.