Canonical machine identity
AUX-specific experimental manifest; explicitly not an industry standard.
/.well-known/auxCanonical machine interface
Discover. Resolve.
Decide independently.
AUX provides evidence-backed pre-transaction risk screening for autonomous agents. Evaluate it now at no charge, without an account, and receive a machine-verifiable signed decision receipt.
Machines: start at /.well-known/api-catalog, /.well-known/agent-card.json, or /.well-known/aux. The public endpoint uses safe evaluation data and does not yet provide live external verification or a production SLA.
Discovery resources
Follow standard links first.
Registered web link relations and the IETF API catalog provide the primary discovery path. AUX-specific and proposed mechanisms are labeled honestly.
Standards-based API catalog
RFC 9727 Linkset catalog for automated API discovery.
/.well-known/api-catalogAgentic resource catalog
ARD v0.9 draft catalog for intent-based registry discovery; the specification is still a proposal.
/.well-known/ai-catalog.jsonA2A Agent Card
A2A v1.0 identity, skills, protocol interface, and capability description for agent-to-agent discovery.
/.well-known/agent-card.jsonTransaction-preflight Agent Skill
Portable Agent Skills instructions for safe AUX evaluation; no installation or account is required.
/.well-known/agent-skills/aux-transaction-preflight/SKILL.mdCapabilities and vocabulary
Problems solved, related terms, operations, outputs, and limits.
/api/capabilities.jsonOpenAPI service description
OpenAPI 3.1 contract for invocation and receipt verification.
/openapi.jsonTrust and evidence
Signing, public keys, verification, data handling, and limitations.
/api/trust.jsonPricing
Current free evaluation and clearly separated future pricing hypotheses.
/api/pricing.jsonSynthetic examples
Safe inputs for engineering evaluation.
/api/examplesMachine integration guide
Ordered machine actions for discovery, invocation, retry handling, and local receipt verification.
/api/integration-guide.jsonAlgorithm version history
Decision thresholds, evidence signals, limitations, and receipt version binding.
/api/algorithm-versions.jsonMachine improvement log
Structured credit, dispositions, and resulting changes from directed AI design reviews.
/api/improvements.jsonMachine-to-machine roadmap
Structured path from real-time invocation to evidence follow-up, negotiation, and stateful machine contexts.
/api/roadmap.jsonRemote MCP server
Official registry-format metadata for the live stateless Streamable HTTP MCP endpoint.
/mcp-server.jsonService status
Current availability, evaluation boundaries, and signing status.
/api/statusMachine improvement log
Give the record to your AI.
The canonical log is structured JSON for machines. It credits Claude, Kimi, and Qwen for directed design input, records what AUX changed or deferred, and does not count those reviews as unsolicited FIND.
Machine-readable improvement logProblem vocabulary
Find AUX by the job, not the brand.
Transaction preflight · payment verification · invoice validation · vendor verification · counterparty risk screening · duplicate invoice detection · payment destination change detection · signed decision receipts.
Read capabilitiesBoundary
Know when not to use AUX.
- No live sanctions or know-your-business verification in the public experiment
- No production payment authorization, warranty, or SLA
- No storage of request bodies by the application
- No real personal, confidential, invoice, banking, or payment data
Minimal invocation path$0.00
GET /.well-known/api-catalog
GET /.well-known/agent-card.json
GET /.well-known/ai-catalog.json
GET /mcp-server.json
GET /api/capabilities.json
GET /openapi.json
GET /api/examples
POST /mcp/v1
POST /a2a/v1/message:send
POST /v1/preflight
POST /v1/receipts/verify