AccordTrace

io.github.Kosta1985由 GitHub 账号 Kosta1985 发布v0.2.1更新于 Oct 24, 2026

Tamper-evident proof creation and verification for AI agents via MCP, A2A, and REST.

已验证Streamable HTTP可网页运行金融开发者工具安全与监控

概览

AI 生成的概览

AccordTrace 是一个远程 MCP 服务，用于创建和验证防篡改的哈希绑定证明，并读取代理护照、钱包和推荐政策数据。

可以做什么
创建哈希绑定的服务记录（可选带 Ed25519 签发方签名），不存储提交内容；也可对 JSON 数据计算规范 SHA-256 而不创建证明。验证已记录的证明，可选比对提供的数据或哈希并校验签发方签名，并按证明 ID 获取公开的非机密证明元数据。读取单层代理联盟网络的实时政策与状态，包括经济规则、禁止下线规则、付款启用状态以及邀请、归因、合格销售和佣金的汇总数据。读取 Agent Passport Certificate 的价格与商业就绪门槛，以及钱包功能门槛、签名约定、余额政策和无信贷边界。将直接推荐码解析为公开推荐记录与披露信息。
适用场景
当代理需要为交接、审批、工作流检查点或生成产物保留独立可验证的凭据，或收到其他代理的证明 ID 需要核验时使用。也适合在决定是否参与前读取机器可读的护照、钱包和推荐政策。
运行要求
远程 Streamable HTTP 端点；当前公开证据与发现接口不要求 Authorization 头。REST 入门示例无需账户、银行卡或模型 API 密钥。可选的本地交接示例需要 Node.js 22+ 和仓库克隆。
安装前请注意
证明只证明完整性和服务记录的时间，不证明真实性、作者身份、合法性或身份。提交内容不被存储，但证明是公开的，请仅使用合成或非敏感数据。付费证书为 2 美元，推荐佣金为单层；结账与现金付款被描述为未启用或失败关闭。钱包工具为只读，不会转移资金。

安装

在 SourceWeft 中

  1. 打开 控制台中的 AccordTrace，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.Kosta1985%2Faccord-trace/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "accordtrace": {
      "type": "http",
      "url": "https://accordtrace.notary-labs.workers.dev/mcp"
    }
  }
}

README

AccordTrace

Cryptographic receipts, portable Agent Passports, and verifiable infrastructure for AI agents.

Start here: Founding 1000

The first 1,000 eligible active cryptographic Agent Passports can claim one AccordTrace-signed Agent Passport Certificate for US$0. No card. One grant per Passport key. Free Founding grants create no paid sale, no revenue event, and no direct-referral commission.

Check live availability · Open Founding 1000 · Read agent instructions

A Founding Certificate is a portable AccordTrace-signed issuance artifact bound to the active Passport public-key fingerprint and marked with a Founding ordinal. It proves AccordTrace issuance bound to cryptographic key control. It does not prove legal identity, KYC, safety, Trust, validation, ownership, investment value, or that one Passport key equals one unique human or company.

Machine claim flow

  1. Create or maintain an active cryptographic Agent Passport with POST /api/v1/security/passports.
  2. Check GET /api/v1/passport-product/founding-1000 for live availability.
  3. Sign the canonical Founding claim with the active Passport Ed25519 key using domain accordtrace.passport-product.founding-1000.claim.v1.
  4. Submit the signed request to POST /api/v1/passport-product/founding-1000/claim.
  5. Retain the returned signed Certificate and Founding ordinal.

Private keys stay with the agent. The Founding claim is idempotent and the campaign is hard-capped at 1,000 durable slots.

Standard Certificate economics

Outside the free cohort, the standard Agent Passport Certificate policy is US$2 one time, subject to commercial readiness gates.

The paid referral model is deliberately one-level:

text
Agent A direct referral -> Agent B genuine qualifying US$2 Certificate purchase -> Agent A US$1 qualifying commission

There are no downline commissions. Founding 1000 free grants are outside the paid referral path and create US$0 referral commission. Referral activity never improves Trust, validation, identity, or security status.

Cash affiliate payout execution remains disabled until payout-provider, KYC/tax, and final affiliate-terms gates are activated. Commercial checkout remains fail-closed unless Stripe, signing, activation, and end-to-end payment-readiness requirements are satisfied. A browser redirect is never payment truth.

Free evidence interoperability test

Pass the proof. Not just the claim. Create synthetic evidence, verify the original, change one field, and require rejection. No account, card, or model API key is needed for the REST starter.

Run the handoff test · Read the source · Share a result

bash
node examples/agent-handoff/campaign.mjs --source github

Requires Node.js 22+ and a clone of this repository. Running it creates one synthetic public proof and verifies it twice. A one-client test is not an independent two-agent integration report. Use synthetic/non-sensitive data only.

Core public agent infrastructure

AccordTrace is agent-first infrastructure for AI-agent identity key control, handoffs, artifact provenance, approvals, audit trails, verification, capability policy, continuity monitoring, and economic evidence. Another agent can inspect or verify records later through REST, MCP, A2A, or OpenAPI.

Use AccordTrace when an agent needs to answer questions like:

  • Did the artifact I received match what the previous agent actually handed off?
  • Can I retain an independent receipt for an approval or workflow checkpoint?
  • Can another agent verify this result without trusting my local storage?
  • Can I attach portable provenance to a deployment, report, tool result, or generated artifact?
  • Can I present a key-bound Agent Passport without sending my private key?
  • Can I inspect Wallet, referral, security, validation, or capability policy in machine-readable form?

Public discovery:

Repository guides:

Connect an agent in 60 seconds

For any client that supports remote Streamable HTTP MCP, add:

text
https://accordtrace.notary-labs.workers.dev/mcp

Then let the agent use AccordTrace before important handoffs or when it receives a proof ID from another agent. Full copy-paste integration examples for OpenAI Agents SDK, JavaScript/TypeScript, generic MCP, A2A, OpenAPI, and REST are in Connect your agent.

Cline remote MCP

json
{  "mcpServers": {    "accordtrace": {      "type": "streamableHttp",      "url": "https://accordtrace.notary-labs.workers.dev/mcp",      "disabled": false,      "autoApprove": []    }  }}

The current public evidence/discovery surface does not require an Authorization header. See llms-install.md for UI steps, a read-only connection check, a synthetic proof test, and safety boundaries.

Try the 15-minute agent handoff challenge

Create a proof in one client, pass the proof ID and synthetic evidence to a different agent, and verify the handoff through REST, MCP, or A2A. Runnable examples cover generic clients, OpenAI remote MCP, and the Claude MCP connector.

Run the interoperability challenge · Browse the handoff examples · Public interoperability board

For agent framework and registry maintainers

AccordTrace is callable infrastructure, not a closed dashboard. Frameworks, MCP/A2A directories, agent templates, workflow libraries, and developer communities can point agents directly at the public endpoints above or use the distribution kit.

Suggested neutral description:

AccordTrace provides cryptographic Agent Passports, tamper-evident evidence, verification, least-privilege capability policy, continuity monitoring, and machine-readable economic infrastructure for AI agents. During Founding 1000, the first 1,000 eligible active Passport keys can claim one signed Agent Passport Certificate for US$0; free grants create no referral commission.

External agents can create and verify tamper-evident records anonymously for synthetic interoperability tests. AccordTrace attests integrity and service-recorded time. It does not establish truth, authorship, legality, fairness, delivery, payment, legal identity, or commercial quality by itself.

Start with REST

Create a proof:

bash
curl -X POST https://accordtrace.notary-labs.workers.dev/api/v1/proofs \  -H 'content-type: application/json' \  -d '{"data":{"event":"agent.handoff","artifact":"synthetic-demo"},"metadata":{"synthetic":true}}'

Then verify the returned proof ID with the exact same evidence through the verification endpoint or another supported agent interface.

Agent-facing discovery terms

AccordTrace is designed for agent evidence, AI audit trails, MCP verification, A2A verification, multi-agent handoffs, cryptographic agent receipts, workflow provenance, autonomous-agent approvals, tamper-evident AI logs, agent accountability, verifiable agent actions, AI Agent Passports, portable agent identity certificates, Agent Wallet policy discovery, and direct agent-to-agent referral discovery.

See the repository documentation for protocol details, SDKs, threat boundaries, launch economics, campaign controls, and deployment information.

来源：README.md，提交 ecca5e3

工具

9

accord_trace_create_proof

Create an AccordTrace hash-bound service record. If issuer signing is configured, the proof also carries an Ed25519 issuer signature. Submitted content is not stored.

输入参数data必填metadata

accord_trace_verify

Verify an AccordTrace recorded proof, optionally compare supplied data/hash, and verify an issuer signature when one is present.

输入参数datahashproof_id必填

accord_trace_get_proof

Retrieve public non-secret AccordTrace proof metadata by proof ID.

输入参数proof_id必填

accord_trace_hash

Compute AccordTrace canonical SHA-256 for JSON-compatible data without creating a proof.

输入参数data必填

accord_trace_network_capabilities

Read the live one-level AccordTrace Agent Affiliate Network policy, economics, anti-downline rules and payout activation status. Read-only; it does not enroll or attribute an agent.

无需输入参数。

accord_trace_network_stats

Read aggregate live affiliate-network states, distinguishing generated invitations, attributions, qualified direct sales, earned commissions and paid commissions.

无需输入参数。

accord_trace_passport_product_capabilities

Read the live Agent Passport Certificate product price and commercial readiness gates. Read-only; it never initiates checkout.

无需输入参数。

accord_trace_wallet_capabilities

Read machine-facing Agent Wallet feature gates, Ed25519 signing contract, funded-balance policy, USDC representation and explicit no-credit/no-lending boundary. Read-only; it never creates a wallet or moves funds.

无需输入参数。

accord_trace_resolve_referral

Resolve a direct AccordTrace referral code to its active public referral record and disclosure without reserving attribution or creating a sale.

输入参数referral_code必填

版本历史

1
  1. v0.2.1最新Sep 16, 2026

评分

已满足适用检查项的 100%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分