Skip to main content
Glama
attestedintelligence

AGA-mcp-server

AGA - 认证治理工件 (Attested Governance Artifacts)

用于 AI 代理和自主系统的加密运行时治理。

npm PyPI License: MIT Tests

# Try it now
pip install aga-governance
python -m aga demo
python -m aga verify demo-bundle.json

功能概述

AI 代理进行的每个工具调用都会通过 AGA 网关。每个调用都会根据策略进行评估,决策(允许或拒绝)将作为已签名的哈希链接治理收据记录下来。收据被收集到证据包中,任何第三方都可以使用标准加密技术进行离线验证。

记录。证明。验证。

Related MCP server: AgentLens

在 Claude Desktop 中使用

添加到您的 Claude Desktop MCP 配置 (claude_desktop_config.json) 中:

{
  "mcpServers": {
    "aga": {
      "command": "npx",
      "args": ["-y", "@attested-intelligence/aga-mcp-server"]
    }
  }
}

Claude 随后可以通过自然语言密封工件、度量完整性、生成证据包并验证合规性。

MCP 工具 (20)

类别

工具

身份

get_server_info, get_portal_state

生命周期

init_chain, attest_subject, revoke_artifact

执行

measure_integrity, measure_behavior, verify_chain

证据

create_checkpoint, generate_evidence_bundle, verify_bundle_offline

隐私

request_claim, list_claims

委托

delegate_to_subagent

审计

get_receipts, get_chain_events

快速入门

验证证据包(3 条命令)

pip install aga-governance
curl -s https://aga-mcp-gateway.attestedintelligence.workers.dev/bundle -o evidence-bundle.json
python -m aga verify evidence-bundle.json

或在浏览器中验证

前往 attestedintelligence.com/verify 并点击“运行验证”。无需安装任何软件。

工作原理

AI Agent                  AGA Gateway                    Verifier
   |                          |                              |
   |-- tools/call ----------->|                              |
   |                    [Evaluate Policy]                    |
   |                    [Sign Receipt]                       |
   |                    [Chain to Previous]                  |
   |<-- PERMITTED/DENIED -----|                              |
   |                          |                              |
   |                    [Export Bundle]                       |
   |                          |--------- evidence.json ----->|
   |                          |                  [Verify Signatures]
   |                          |                  [Verify Chain]
   |                          |                  [Verify Merkle Tree]
   |                          |                  [PASS / FAIL]

MCP 治理代理

将 AGA 作为任何 MCP 客户端和任何 MCP 服务器之间的透明代理运行。每个工具调用都会根据策略进行评估并生成签名收据。

# Start the proxy with an upstream MCP server
npx tsx src/proxy/index.ts start --upstream "npx -y @modelcontextprotocol/server-filesystem /tmp/test" --profile standard

# Export the evidence bundle
npx tsx src/proxy/index.ts export --output evidence.json

# Verify
npx tsx src/proxy/index.ts verify evidence.json

该代理拦截 tools/call 请求,根据密封的策略工件对其进行评估,并生成签名收据。被允许的调用将转发到下游服务器。被拒绝的调用将返回 MCP 错误。每个决策都通过哈希链接到一个防篡改链中。

内置三种策略配置文件:

  • permissive - 记录所有内容,不拦截任何内容(默认)

  • standard - 速率限制 + 拦截破坏性操作

  • restrictive - 明确的工具允许列表,拒绝所有未知工具

验证(5 个步骤)

  1. 算法检查 - 证据包声明使用 Ed25519-SHA256-JCS,其他任何算法均会导致失败

  2. 收据签名 - 基于 RFC 8785 规范化 JSON 的 Ed25519 签名(排除签名字段)

  3. 链完整性 - 每个收据的 previous_receipt_hash = 前一个收据的 SHA-256 哈希值

  4. Merkle 证明 - 遍历兄弟节点/方向至根节点,与证据包根节点进行比较

  5. 证据包一致性 - 证明计数 = 收据计数,叶子哈希与收据哈希匹配

加密原语

原语

用途

Ed25519

收据签名

SHA-256

哈希链、Merkle 树、叶子计算

RFC 8785 (JCS)

用于确定性签名的规范化 JSON

Merkle 树

将所有收据绑定到单个可验证的根

实时网关

演示网关部署在 Cloudflare Workers 上:

# Check status
curl https://aga-mcp-gateway.attestedintelligence.workers.dev/health

# Export evidence bundle
curl https://aga-mcp-gateway.attestedintelligence.workers.dev/bundle -o evidence-bundle.json

Python SDK

pip install aga-governance
from aga import AgentSession

with AgentSession(gateway_id="my-gateway") as session:
    session.record_tool_call(
        tool_name="search_web",
        decision="PERMITTED",
        reason="tool in allowlist",
        request_id="req-1",
    )
    bundle = session.export_bundle()
    result = session.verify()
    assert result["overall_valid"]

测试套件

TypeScript 和 Python 中超过 355 个自动化测试:

  • TypeScript MCP 服务器: 218 个测试 (vitest)

  • Python SDK: 137 个测试 (pytest)

  • 跨语言测试向量: 9 个类别中的 37 个向量

npm test                              # TypeScript tests

对于 Python SDK,请从 PyPI 安装 aga-governancehttps://pypi.org/project/aga-governance/

项目结构

src/                   # Core protocol: artifacts, receipts, chain, Merkle, crypto, portal state machine
  core/                # Governance primitives (artifact, receipt, chain, portal, bundle)
  crypto/              # Ed25519, SHA-256, BLAKE2b, Merkle, JCS canonicalization
  proxy/               # MCP governance proxy (transparent interception + policy enforcement)
  tools/               # MCP tool handlers (20 tools)
  middleware/          # Zero-trust governance enforcement wrapper
independent-verifier/  # Standalone verifier with zero AGA imports
scenarios/             # Deployment scenarios (SCADA, drone, AI agent)
tests/                 # TypeScript test suite (218 tests)

链接

安全性

请参阅 SECURITY.md 以了解漏洞报告。

贡献

请参阅 CONTRIBUTING.md 以了解开发设置和指南。

许可证

MIT


Attested Intelligence Holdings LLC

Install Server
A
license - permissive license
B
quality
A
maintenance

Maintenance

Maintainers
Response time
5dRelease cycle
2Releases (12mo)
Commit activity

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Servers

  • A
    license
    A
    quality
    A
    maintenance
    Local zero-trust permission gateway for AI agents. Enforces policy-based tool authorization, human approvals, scoped permissions, and cryptographically verifiable audit logs.
    4
    5
    Apache 2.0
  • A
    license
    A
    quality
    A
    maintenance
    AI-agent observability server whose distinguishing feature is a SHA-256 hash-chained, tamper-evident audit log with chain verification and signed export. Works with Claude Desktop, Cursor, and any MCP client.
    22
    2
    20
    MIT
  • F
    license
    A
    quality
    C
    maintenance
    MCP server that auto-emits tamper-evident receipts for every tool call, enabling EU AI Act Article 12 compliance with signed, chain-linked receipts.
    1
  • A
    license
    B
    quality
    A
    maintenance
    A governance proxy for AI tools — every MCP/agent tool call is policy-gated, secret-redacted, and written to a hash-chained, offline-verifiable audit trail.
    13
    MIT

View all related MCP servers

Related MCP Connectors

  • Pre-action attestation perimeter for AI agents — 8 primitives, signed C18 receipt per call.

  • Hash-chained HMAC-signed audit log MCP for A2A (agent-to-agent) calls. Every tool-call, agent-ha...

  • Bitcoin-anchored, tamper-evident audit log for AI agents — record, disclose and verify actions.

View all MCP Connectors

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/attestedintelligence/aga-mcp-server'

If you have feedback or need assistance with the MCP directory API, please join our Discord server