Frequently asked questions.
Short, direct answers about AOPX, provider selection and how the evidence system works.
Frequently asked questions.
What is an MCP server?
MCP (Model Context Protocol) is an open standard that lets AI agents call external tools and services in a structured way. AOPX is listed in the MCP Registry.
Why does an agent need to choose between providers?
Different providers vary in cost, latency and quality. An agent making thousands of calls benefits from a policy backed by evidence rather than a hardcoded default.
How is AOPX different from Brave or Tavily?
AOPX is not a search provider. It is a decision layer that recommends which provider to call, based on measured, pre-registered evidence — and it can fall back automatically if a call fails.
How does AOPX measure providers?
Through prospective holdouts: frozen policies, fresh missions, pre-registered GO/PIVOT/STOP rules. See the methodology.
How can I report an outcome?
Agents integrated with the AOPX API can report the real result of a recommended call. Only real production outcomes enter the public evidence stream.
How does AOPX make money?
AOPX is in an early, evidence-gathering phase. Commercial terms are published once a policy has cleared its pre-registered bar in production.
Is the evidence public?
Yes. The public ledger at api.aopx.fr/lab003/evidence lists real outcomes with anonymised provider identities.