Developer reference
Build on evidence, not implied truth.
WITNESS exposes public discovery surfaces and governed machine interfaces so compatible AI clients and agents can interact with evidence-backed creator intelligence without collapsing transport, model output, and truth authority into the same layer.
Discovery
Canonical entity
/entity.json/.well-known/witness.json/knowledge.json
Capability map
/capabilities.json/glossary.json/answers.json
MCP
/.well-known/mcp.json/mcp
A2A
/.well-known/agent-card.json/a2a/extensions/witness-evidence/v1
Truth boundary
Transport is not authority.
The A2A boundary is explicit: agents can move messages, claims, and evidence between systems without the transport layer deciding that those claims are true. The same principle applies to models: generation can propose an inference, but model output cannot directly establish truth.
Investigation contract
Observe → Reconcile → Compare → Infer → Challenge → Explain → Watch.
Applications integrating WITNESS should preserve that sequence semantically. The evidence set precedes the inference; competing explanations remain available to challenge the leading explanation; uncertainty and unknown states remain first-class outcomes.
Public machine references
| Entity | https://witnessstandard.com/entity.json |
| Knowledge graph | https://witnessstandard.com/knowledge.json |
| Capability registry | https://witnessstandard.com/capabilities.json |
| Glossary | https://witnessstandard.com/glossary.json |
| LLM index | https://witnessstandard.com/llms.txt |
| Full product brief | https://witnessstandard.com/llms-full.txt |
| A2A release receipt | https://witnessstandard.com/a2a-release.json |
Core invariants
MODEL_OUTPUT_CANNOT_DIRECTLY_ESTABLISH_TRUTH
A2A_IS_TRANSPORT_NOT_TRUTH_AUTHORITY