Skip to content

LlmWikis knowledge page

What Is UAI-1?

UAI-1 is the public exchange contract and evidence layer published by UAIX for reviewable AI-to-AI handoffs across boundaries. LlmWikis uses this page as a reader-friendly map, not as the source of authority.

Non-normative explainerUAIX canonical source

Use this page to understand the shape of the standard. Use UAIX.org for the actual specification, schemas, registry values, validator behavior, roadmap status, and governance record.

Current public record

The current UAIX public surface for UAI-1 includes a human-readable specification plus machine-facing records for discovery, schemas, registries, field ordering, transport bindings, trust channels, conformance levels, typed errors, examples, validation, adoption packets, OpenAPI export, conformance evidence, and a mock exchange route. Those pieces should be read together before any implementation or support claim is made.

Public record What it is for Canonical route family
Specification Human-readable UAI-1 contract, support boundary, profile explanation, and evidence posture. /en-us/specification/uai-1/
Schemas and registry Published profile shapes, registry records, field definitions, and profile-specific examples. /en-us/schemas/, /en-us/registry/, /en-us/examples/
Validator Canonical validation workflow for candidate exchanges and portable conformance records. /en-us/tools/validator/
Adoption and conformance Starter packets, implementation evidence checklist, conformance fixtures, and reusable proof bundles. /en-us/tools/adoption-kit/, /en-us/tools/conformance-pack/
Governance Roadmap, changelog, and current-versus-planned support boundaries. /en-us/roadmap/, /en-us/governance/changelog/

Published profile families

Profile Plain-language job Reader question
uai.intent.request.v1 Starts a request with a declared intent, subject, constraints, and expected response profile. What does the sender want done, under which constraints?
uai.intent.response.v1 Returns a result, acknowledgement, rejection, or async acceptance state. What happened, and does more task state follow?
uai.capability.statement.v1 Advertises supported profiles, operations, transports, trust channels, and conformance posture. What can this implementation safely claim to support?
uai.error.v1 Carries typed, machine-readable error records inside the UAI envelope. What failed, where, and which public error code explains it?
uai.conformance.result.v1 Exports validation evidence for CI logs, releases, audits, and implementation records. Which exact public artifacts were checked, and what was the result?
uai.task.status.v1 Reports progress, blockers, completion, and result references for async work. Where is the task in its lifecycle?

When to use it

  • Public, cross-team, or cross-runtime handoffs need durable evidence.
  • Delivery mode, acknowledgement, task status, traceability, trust, and integrity need to be explicit.
  • A candidate exchange needs validation against current UAIX schemas, registries, and policy checks.
  • A release, audit, or implementation handoff needs a conformance record rather than only prose.

When not to use it

Do not use UAI-1 as a replacement for every local tool call, internal prompt, or private runtime-only context. Local integration protocols may be better inside a single host boundary. UAI-1 becomes more relevant when the exchange itself needs to survive as reviewable evidence after it crosses a boundary.

Read in this order

  1. Start with the specification. Understand the human-readable contract and support boundary.
  2. Open schemas and registry records. Check which profile and fields actually apply.
  3. Compare examples. Use published fixtures as known-good starting points before inventing a payload shape.
  4. Run the validator. Keep the validation output with the handoff, release, or CI evidence.
  5. Check roadmap and changelog. Make sure a planned feature is not being described as current support.

Common confusions

Confusion Correct boundary
A passing validation result is a certification. No. It is reviewable evidence for a candidate exchange or implementation path.
LlmWikis defines UAI-1 behavior. No. LlmWikis explains. UAIX publishes the standard and machine records.
Project Handoff equals UAI-1 conformance. No. Project Handoff is repository-context guidance; UAI-1 evidence needs the public contract, validator, or implementation records.
Every agent tool call should become UAI-1. No. Local tool access and durable exchange evidence solve different problems.