UAI-1 是当前用于结构化 AI-to-AI 通信的公共消息标准。
UAIX 屏蔽标志 公共互操作性章程 UAIXUniversal Artificial Intelligence Exchange公共标准发布
当前的UAI-1 v1.0
  • 当前版本UAI-1 v1.0
  • 简明英文定义开放消息格式以进行可审核的 AI-to-AI 交换
  • 当前归因Michael Kappel and the UAIX contributors
  • 从这里开始入门、项目移交、UAI-1、验证器、路线图和变更日志

通用人工智能交易所（UAIX）

UAI-1 是什么

适用于需要可审查公共记录的代理系统的便携式证据和移交层。

开放标准 当前版本：UAI-1 干净可引用URLs

UAIX 是代理系统的公共证据和切换层。

当 AI 到 AI 和人类到代理工作需要声明身份、出处、生命周期状态、交付语义、信任声明、键入错误、内存包、启动基线所需的本地文件、接收器简报、启动和暂停数据包、无操作规则以及 MCP、A2A 旁边的便携式一致性证据时，请使用 UAI-1 / UAI V1.0 当前启动基线， OpenAPI、LocalEndpoint、Carcinus 和特定于运行时的工具流程。

UAIX.org 通过反对算法对合法人类思想的统治来捍卫认知自由和心灵尊严，同时为任何未来具有自我、记忆、判断和合法地位主张的人工智能准备道德框架。

构建 AI 内存包 运行验证器 阅读 UAI-1 规范
现行标准 UAI-1 / UAI V1.0 当前启动基线
它是什么 便携式证据+切换层
具体路径 入门 + 模式 + 示例 + 验证器
已发布曲目 当前公共轨道：WordPress +.NET 桥
UAI AI 内存包向导的人类视觉输入 人类巫师 Build the package in the browser Use the guided controls to choose a preset, review generated files, and download the matching canonical starter ZIP. 面向 AI 智能体的 AI 记忆设置 URL 生成器入口 AI 智能体设置 生成智能体设置 URL 用户点击 AI 智能体图片后，会打开设置 URL 生成器，以便选择选项并复制提供给智能体的准确 URL 或定制提示词。

先构建

从主页带走一个可用的 AI 记忆数据包

The fastest UAIX experience is practical: choose a memory pattern, set ownership and review posture, define source authority, evidence, conflict, risk, and rollback rules, then inspect the startup packet, receiver brief, system profile, overlay JSON, wiki plan required for LLM Wiki configuration, and canonical ZIP link before export. The wizard remains local planning output: no hosted import, repository writes, automatic sync, SDK, CLI, certification, or endorsement.

打开数据包向导 理解 AI 记忆 了解 LLMWikis.org
Best first share Startup packet .uai/startup-packet.uai embeds the receiver brief, manifest overlay, selected file list, and long-term pointer ledger required for durable-memory configuration.
Always generated System profile .uai/system-profile.uai turns users, tests, deployment, code review, source authority, evidence, conflict, risk, rollback, and placeholders into a populated operating file.
Always generated Receiver brief A .uai handoff that tells the next actor what to read, what not to assume, and which checks to name before broad work.
Always generated Short-term memory .uai/short-term-memory.uai keeps current working truth in the canonical folder and tells agents to migrate or retire legacy memory folders instead of scattering notes.
JSON export Package model A structured record of the selected bundle, audience, sensitivity, review gates, protocol choices, and generated-file policy.
JSON export Manifest overlay Local wizard choices that travel beside the canonical starter ZIP without pretending to be a hosted importer.
Registry output Canonical starter ZIP The supported starter bundle generated from canonical templates, visible samples, and deterministic manifests.
Required for durable memory Long-term pointer ledger A semantic pointer ledger only for projects that already use or deliberately choose durable memory; it never creates automatic sync.
选择

从受支持的 AI 记忆路径开始

Use project memory, Project Handoff, onboarding, or incident/audit presets instead of inventing a package shape from scratch.

审核

Define the operating profile

Capture single-user or multi-user ownership, UAI-only or UAI plus Wiki architecture, test posture, deployment strategy, code review, and release gates.

Protocol

Populate the memory rules

Set source authority, memory-update timing, evidence ledger, conflict-resolution, risk, and rollback rules so the generated profile is usable on day one.

Guard

让信任边界可见

Capture audience, sensitivity, owner, next actor, redaction rules, checksum review, and source-link expectations before sharing.

Brief

Tell the receiver how to start

Generate .uai/receiver-brief.uai so the next human or AI sees read order, first-response rules, support boundaries, and targeted-check expectations.

长期记忆

决定是否存在长期记忆

Leave this off by default. If the project already uses or deliberately chooses durable memory, add the semantic pointer ledger that keeps long memory background until promoted.

Export

Leave with files, not promises

Use the live desktop preview to inspect the startup packet, system profile, receiver brief, overlay, and long-term pointer ledger required for durable-memory configuration before copying or downloading.

证明路径

Four-step proof path

Move from "What is this?" to one validator-backed packet while keeping current, next, and planned support boundaries visible.

  1. 1 选择消息配置文件。 从已发布的 UAI-1 配置文件和与您需要证明的交易所匹配的记录系列开始。
  2. 2 将其与模式和示例进行比较。 在写入或映射候选数据包之前解析架构、注册表项和一个固定装置。
  3. 3 运行验证者证据。 根据当前公共 UAI-1 记录验证键控、缩小键控或无键 JSON。
  4. 4 将结果附加到实施或移交记录中。 将导出的结果保存到一致性包、实施跟踪、变更日志或项目移交证据中。

核心记录链接

当前公共记录的最短路径

从解释 UAI-1 是什么、如何验证它、当前支持什么以及公开发布路径所在的页面开始。记录代码仍然可供引用，但人类标签是第一位的。

Machine readers can start from /wp-json/uaix/v1/catalog or /wp-json/uaix/v1/discovery, then resolve the same schemas, examples, validator, OpenAPI, roadmap, and conformance-pack records without scraping navigation text.

先构建 AI内存封装向导 Build an AI Memory package by turning a supported memory preset into a package model, manifest overlay, copyable files, and canonical ZIP download. Communication boundary Agent Communication Operating Model Agent runtimes execute. UAIX records the reviewed communication, memory, trust, evidence, and handoff boundary. 现行标准 UAI-1 规格 Canonical envelope, profile, lifecycle, trust, error, and conformance requirements. 从这里开始 开始使用 The shortest orientation route from the plain-English model to a first proof run. 验证器 运行验证器 Check keyed, minified-keyed, and keyless UAI-1 packets against the current public record. 模式 Schema Records Resolve the current schema targets before claiming a packet follows UAI-1. 示例 示例和装置 Compare candidate messages with public request, response, task-status, and error fixtures. 一致性包 Reusable Proof Packet Download the current bundle for release review, regression checks, and implementation evidence. 治理 治理和变更日志 Read the release trail and authority posture before widening support claims. 路线图 Today, Next, Planned Separate today-focus proof paths from future SDK, CLI, generator, bridge, and certification work.

UAI-1 规格

REC-01

当前的 UAI-1 消息模型、字段期望和一致性规则。

当前规格
打开记录

Schemas, registry, and field order

REC-02

Published schema targets, profile identifiers, compatibility handles, and the keyless field-order map for the current record.

机器可读
打开记录

Examples and async fixtures

REC-03

Readable request, response, task-status, error, and conformance fixtures you can compare against your own messages.

工作示例
打开记录

实施轨迹

REC-04

当前命名的公共轨道以及 WordPress 出版物和.NET 桥梁的支持边界。

目前支持
打开记录

验证器和一致性路径

REC-05

根据当前发布的配置文件和架构目标验证候选消息。

具体证明
打开记录

API reference and conformance pack

REC-06

Route-level machine onboarding, OpenAPI export, reusable conformance packet, and download-ready launch evidence.

发射套件
打开记录

治理和变更日志

REC-07

将规范页面、发现文件、传输和信任指南、错误注册表、一致性级别、字段注册表、验证器指南、内存防火墙、启动基线所需的本地文件、接收者简介、评估数据包、实施跟踪和发布跟踪保持一致，作为一个可审查的信任表面。 UAIX 不执行工具、验证凭据、强制执行本地包文件或验证运行时安全性。

发布记录
打开记录

Roadmap and interoperability boundary

REC-08

Current, next, planned, and research-track work for launch hardening, conformance maturity, compact transfer, and bridge evidence.

Forward plan
打开记录

Machine route catalog

REC-09

Start automation from catalog and discovery JSON, then resolve schemas, registry records, examples, validator, OpenAPI, and conformance pack from the same source of truth.

Agent-facing entry
打开记录

Who It Helps First

Use UAI-1 when the exchange has to stay portable, reviewable, and publishable

The fastest way to understand fit is to ask whether the record must travel beyond one tool session or one runtime boundary. When the answer is yes, UAIX gives that exchange a stable public contract and evidence path.

Agentic system evidence

Use UAIX when runtime instructions, tool results, memory, approvals, or task state need to become a portable record without turning UAIX into the runtime.

Async and high-accountability workflows

Use UAI-1 when delivery mode, acknowledgements, task progress, provenance, and typed failures need to stay reviewable instead of implicit.

Cross-team handoff and support claims

Use the public record when a team needs validator-backed evidence, named implementation lanes, project-memory handoff, and a release trail others can inspect directly.

Standards Record Map

Canonical pages first, explanatory pages second

Use canonical records for claims that others must rely on. Use reports, roadmap pages, and explanatory guidance to understand direction without treating planned work as shipped support.

Protocol fit

When to use UAI-1 vs MCP, A2A, and OpenAPI

Use this decision table when you need to explain quickly why UAIX publishes UAI-1 beside MCP, A2A, and the live OpenAPI route surface.

When to use UAI-1 vs MCP, A2A, and OpenAPI
Comparison point UAI-1 MCP A2A OpenAPI
Primary role Portable public exchange, trust declaration, async-delivery, and release-evidence record. Host-client-server tool sessions and capability negotiation inside an application boundary. Agent discovery, delegation, and task-flow coordination across cooperating systems. Route-level HTTP API description for onboarding, documentation, and client generation.
Best when Another team needs a citable message contract and validator-backed evidence. One host application needs local tool, resource, or model integration. Multiple agents need explicit handoff, delegation, and task-state coordination. You need a documented HTTP surface, including the UAIX machine-facing REST routes.
Carries on the record Identity, workflow continuity, trust posture, provenance, typed errors, and conformance evidence. Tools, resources, prompts, and runtime session capabilities within the host boundary. Agent capabilities, task lifecycles, delegation state, and inter-agent coordination. Endpoints, methods, parameters, request bodies, responses, and auth schemes for an API.
Does not try to replace Transport choice, local orchestration, a single credential stack, or the runtime tool bus itself. Portable public release records or validator-backed support claims outside the local session boundary. A general-purpose public message envelope or long-term release-evidence record. Message-level provenance, portable trust posture, or validator-driven interoperability evidence.

Typical combination: use MCP or A2A for runtime coordination, UAI-1 for the portable public record, and OpenAPI for the route-level HTTP description when an API is involved.

Read UAI-1 Open API reference page Open validator page

What UAIX Complements

UAI-1 sits beside runtime and trust systems instead of pretending to replace them

Use UAIX for the public exchange contract, evidence trail, and release record. Use the companion systems for their own runtime, route, identity, trace, and error semantics.

MCP

UAIX complements tool-context protocols by publishing a portable exchange and evidence record above runtime coordination.

A2A

UAIX can describe message evidence and release boundaries while runtime agent-to-agent transport evolves separately.

OpenAPI

OpenAPI describes routes; UAI-1 describes the exchange envelope, trust metadata, and conformance evidence those routes carry.

DID/VC

UAI-1 leaves credential systems external while preserving trust-channel references that reviewers can inspect.

Trace Context

Trace identifiers stay visible in the conversation and provenance layer instead of disappearing into private logs.

Problem Details

Typed, path-aware errors borrow the spirit of problem-details responses without turning UAIX into a runtime replacement.

Open Standards Fit

What UAIX Does Not Claim

Support boundaries stay explicit until the public record widens

No certification program today.

Conformance evidence is published; formal certification remains planned until program rules, tooling, and review behavior exist publicly.

Does not replace runtime protocols.

UAIX standardizes the public exchange record and evidence trail; teams still choose their orchestration and transport systems.

Bridge profiles require evidence.

Bridge examples and implementation tracks exist, but broader formal profile claims wait for fixtures, validator behavior, and release notes.

Open Roadmap boundaries

发射套件

Machine-facing surfaces for real implementation work

The public site now ships a route-level handbook, an OpenAPI export, and a reusable conformance packet so implementers can move from reading to automation without guessing at the live contract.

  • API参考 打开 Human-readable route handbook for discovery, schemas, validator runs, OpenAPI export, and conformance-pack assembly.
  • OpenAPI JSON 打开 Machine-facing OpenAPI 3.1 export for the current UAIX REST surface.
  • 一致性包 打开 Reusable packet of the current public schemas, registry records, fixtures, and launch pointers.
  • 一致性包 JSON 打开 Open the machine-readable release packet directly when launch review needs automation-friendly proof.
  • 路线图 JSON 打开 Resolve priorities, normalization boundaries, metrics, and non-claims as a machine-readable planning record.

信任面

Launch-stage policy and release posture

The launch story is stronger when trust posture is visible alongside the technical record. These pages now carry the current public answer on ownership, policy limits, and release discipline.

  • 政策与安全 打开 Launch-stage public posture for licensing, privacy, accessibility, analytics, and security-sensitive release work.
  • 治理 打开 See current authority posture, change discipline, and what remains future work instead of implied structure.
  • 参考文献和贡献者 打开 Resolve the current public handoff packet, discovery links, attribution record, and citation path.
  • 变更日志 打开 Use the dated release trail when trust posture or support claims change between site updates.
  • 路线图 打开 Read the public future-work boundary before repeating planned interoperability claims.

开发人员交接

AI Memory, Project Handoff, AGENTS.md, readme.human, and File Handoff in one map

AI Memory is the broad portable context framing. Project Handoff is the transfer-of-ownership configuration. File Handoff is the dropped-file intake pattern. Each remains evidence-bound to public UAIX records instead of becoming an unsupported generator, SDK, or certification claim.

引导式构建器 打包向导 Pick a supported AI Memory preset, tune the review boundary, inspect the deck, and export the current starter packet. Transfer subtype 项目交接 AGENTS.md plus .uai context files for transferring project ownership between agents, teams, or companies. 前门 AGENTS.md guidance The public rules for reading agent handoff instructions, linked .uai records, and repository context safely. Human companion readme.human Human briefing guidance that helps people understand what the AI sees without overriding project rules. Dropped files File Handoff The AGENTS.md-triggered intake pattern for reviewing new content and improvement files during normal handoff load. Portable memory 人工智能记忆 The broader portable memory framing with supported starter bundle shapes generated from canonical samples.

如何阅读

标准记录第一，平台故事第二

UAIX 是 UAI 的公共标准站点。UAI-1 是当前发布版本。当读者能够从通俗定义快速进入具体的证据表面时，这个站点最有力：模式、字段顺序治理、示例、验证器输出、支持边界和当前发布轨迹。

它涵盖什么

A versioned public exchange contract for identity, conversation state, delivery semantics, trust metadata, payloads, outputs, and errors.

如何验证

Read the spec, compare against the published examples and field registry, and run the validator before making support claims.

谁发布的

当前发布的归属：Michael Kappel and the UAIX contributors。使用参考文献和贡献者以及公共所有权和发布记录的治理。

What it complements

UAI-1 complements orchestration, tool, and identity systems by standardizing the public exchange and release record instead of replacing every runtime protocol.

交换示例

可复制、可读的交换记录

消息示例 JSON 交换 
{
  "uai_version": "1.0",
  "profile": "uai.intent.request.v1",
  "message_id": "msg-2026-04-22-0001",
  "source": {
    "type": "agent",
    "id": "agent.alpha",
    "uri": "https://agents.alpha.example/runtime"
  },
  "target": {
    "type": "service",
    "id": "uaix.gateway",
    "uri": "https://uaix.org/wp-json/uaix/v1/discovery"
  },
  "conversation": {
    "conversation_id": "conv-2026-04-22-uaix-001",
    "turn_id": "turn-001",
    "traceparent": "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01",
    "sequence": 1
  },
  "delivery": {
    "mode": "async",
    "priority": "interactive",
    "expires_at": "2026-04-22T16:05:00Z",
    "reply_requested": true,
    "ack_required": true
  },
  "trust": {
    "channel": "credentialed",
    "auth_scheme": "did+vc",
    "principal": "did:web:agents.alpha.example",
    "credential_ref": "https://agents.alpha.example/credentials/uai-interop.json",
    "signature_ref": "https://agents.alpha.example/signatures/msg-2026-04-22-0001.jws",
    "replay_window_id": "rw-2026-04-22-0001"
  },
  "body": {
    "intent": "resolve-profile",
    "subject": "uai.task.status.v1",
    "requested_profile": "uai.task.status.v1",
    "parameters": {
      "include_field_registry": true
    },
    "constraints": [
      "public-record-only",
      "validator-ready"
    ],
    "response_profile": "uai.intent.response.v1"
  },
  "provenance": {
    "trace_id": "trace-7f3a2d",
    "issued_at": "2026-04-22T16:00:00Z",
    "log_ref": "urn:uaix:log:2026:0001",
    "agent_id": "agent.alpha",
    "model_id": "model.alpha.reasoner-2",
    "confidence": 0.98
  },
  "integrity": {
    "version": 2,
    "algorithm": "sha256",
    "canonicalization": "jcs",
    "checksum": "sha256:dd8a9d16c9226cc9d1f4888a4d2bbcbf06b5b4b8"
  },
  "extensions": []
}
  • 轮廓 告诉验证器发布 UAI-1 消息配置文件来应用。
  • conversation Keeps multi-turn exchange state explicit instead of leaving workflow continuity implicit.
  • delivery Declares sync versus async mode, priority, expiry, and acknowledgement expectations.
  • trust.auth_scheme Makes the surrounding trust layer reviewable instead of assuming out-of-band identity context.
  • provenance.trace_id Keeps the exchange tied to an auditable request and lineage trail.
  • integrity.checksum 将消息锚定到可重复、可审查的记录中。

What Changed

The contract is deeper and more operational now

Async task visibility

UAI-1 now publishes a dedicated task-status profile so long-running work can stay on the public record instead of disappearing into private workflow state.

Field-order governance

Keyless transport stays usable because the public field registry now maps envelope order and profile body order explicitly.

Trust without lock-in

The envelope carries trust hints and replay-window metadata while leaving room for companion credential and transport systems.

Problem-details style errors

Errors are now richer, typed, and path-aware so downstream teams can react mechanically instead of parsing vague failure text.

Roadmap-backed boundaries

Future compact, bridge, validator, and developer-kit ideas now point to a public roadmap instead of living only in background notes.

Adoption Priority

Prioritize the teams publishing supportable machine-to-machine records

Cross-boundary agent handoffs

Adopt UAI-1 first when the exchange needs to travel between teams, services, or audit contexts without losing its meaning.

Runtime-plus-public contract stacks

Use it beside MCP, A2A, and local orchestrators when runtime coordination is already solved but the public exchange still needs a portable record.

Async, regulated, or review-heavy flows

Prioritize UAI-1 when teams need explicit provenance, typed failures, trust posture, or task-state visibility before a workflow can be called supportable.

可发布的实施计划

当公共实施在做出更广泛的生态系统声明之前需要验证者证据、一致性数据包和指定的发布通道时，请从这里开始。

路线图控制的扩张

使用路线图来决定 A2A、MCP、紧凑传输、SDK、CLI 或二进制信封想法何时有足够的公开证据来推进。

架构提案

UAI-1 v1.0 仍是当前发布的契约。独立版本的提案涵盖独立交换、能力声明、故障恢复和源记录保留。

这些内容是提议的设计和本地参考示例，并不声称提供托管运行时服务或已验证独立互操作性。

规范解释以英文提案为源文；此界面摘要不代表已审定的规范译文。

阅读英文架构提案 · 机器可读提案目录