Trust layer
Why Agntbase does not fake MCP or OAuth signals
A higher readiness score is tempting. But publishing machine-readable metadata for systems that do not exist makes the web less trustworthy. Agent readiness should be honest.
A higher readiness score is tempting. But publishing machine-readable metadata for systems that do not exist makes the web less trustworthy. Agent readiness should be honest.
If a scanner asks for OAuth discovery, MCP server cards, API catalogs and skills indexes, teams can be tempted to create empty files just to turn red checks green.
That is the wrong incentive. An agent may rely on those files. If the endpoint claims a capability that is not real, the agent wastes time, fails a workflow, or learns that the source is unreliable.
Link headers, API catalog, Agent Card, skills index, bot hints and content preferences when they point to real resources.
No fake OAuth metadata if there is no auth server. No fake MCP server card if there is no real MCP server.
Trust improves when the site says what is live, what is planned and what still needs human confirmation.
A scanner is useful because it reveals missing infrastructure. But the actual product is reliable interpretation: can an AI system understand the business, cite the source, respect the limits and route the next action safely?
Agntbase would rather show an honest 4/6 for API/Auth/MCP discovery than a fake 6/6 that creates brittle machine behavior.