IAGA-Sentinel 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
IAGA Sentinel sits next to your AI agents and answers the one question the agent无法回答。突出其AI工作流的价值和优势。
IAGA-Sentinel 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
IAGA Sentinel sits next to your AI agents and answers the one question the agent无法回答。突出其AI工作流的价值和优势。
IAGA-Sentinel 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:cargo install(推荐) cargo install iaga-sentinel # 方式二:从源码编译 git clone https://github.com/EdoardoBambini/IAGA-Sentinel cd IAGA-Sentinel cargo build --release # 二进制在 ./target/release/iaga-sentinel
# 查看帮助 iaga-sentinel --help # 基本运行 iaga-sentinel [options] <input> # 详细使用说明请查阅文档 # https://github.com/EdoardoBambini/IAGA-Sentinel
# iaga-sentinel 配置说明 # 查看配置选项 iaga-sentinel --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export IAGA_SENTINEL_CONFIG="/path/to/config.yml"
<p align="center"> <strong>The EU AI Act conformity evidence layer for AI agents.</strong> </p>
<p align="center"> Cryptographically signed, replay-verifiable, EU-sovereign proof of every action an agent takes, mapped to AI Act Article 12 and Annex IV. </p>
<p align="center"> <img src="https://img.shields.io/badge/version-1.5.1-0f9d6b?style=flat-square" alt="version" /> <img src="https://img.shields.io/badge/license-BUSL--1.1-0f9d6b?style=flat-square" alt="license" /> <img src="https://img.shields.io/badge/EU%20AI%20Act-Art.%2012%20and%20Annex%20IV-0B0F0E?style=flat-square" alt="EU AI Act Article 12 and Annex IV" /> <img src="https://img.shields.io/badge/tests-275%20passing-0f9d6b?style=flat-square" alt="tests" /> <img src="https://img.shields.io/badge/Rust-stable-0B0F0E?style=flat-square" alt="Rust" /> </p>
<p align="center"> <a href="#what-iaga-sentinel-is">What IAGA Sentinel is</a> · <a href="#eu-ai-act-mapping">EU AI Act mapping</a> · <a href="#quickstart">Quickstart</a> · <a href="#features">Features</a> · <a href="#architecture">Architecture</a> · <a href="#documentation">Docs</a> · <a href="#status">Status</a> </p>
<p align="center"> <img src="media/iaga-sentinel-promo.gif" alt="IAGA Sentinel, signed tamper-evident audit for AI agents" width="760" /> </p>
---
Cargo features on iaga-sentinel-core:
| Feature | Default | Adds |
|---|---|---|
sqlite | ✅ | SQLite backend for audit + receipts. |
postgres | ❌ | Postgres backend. |
receipts | ✅ | Ed25519-signed Merkle-chained receipts (M2). |
apl | ✅ | Agent Policy Language parser + evaluator + iaga policy ... (M3). |
reasoning | ✅ | Reasoning plane scaffold + iaga reasoning info (M3.5). |
ml | ❌ | tract-onnx ML backend; opt-in, +~5 MB binary, +~2 min cold compile. |
kernel | ✅ | Enforcement kernel + iaga run + iaga kernel status (M4). |
linux-bpf | ❌ | Linux eBPF/LSM scaffold + ringbuf API. Real Aya-rs loader lives in IAGA Sentinel Enterprise. |
plugin-attestation | ❌ | Offline Sigstore bundle + CycloneDX SBOM verify + iaga plugins verify (1.2). |
apl-wasm | ❌ | APL to WASM codegen MVP + iaga policy compile (1.2). The Hindley-Milner type checker (iaga policy check) is always on, no feature needed. |
otel-receipts | ❌ | Emit each signed receipt as an OpenTelemetry span on /v1/telemetry/spans and /v1/telemetry/export, so any OTel stack ingests the evidence. The span includes iaga.receipt.id, iaga.chain.head, iaga.policy.verdict, and iaga.is_authoritative. No new dependency. |
plugin-manifest-signing | ❌ | Ed25519-signed plugin manifests verified at load against trusted keys, plus iaga plugins sign-manifest and verify-manifest (1.3). Orthogonal to plugin-attestation. |
default = ["demo", "sqlite", "receipts", "apl", "reasoning", "kernel"].
The standalone verifier iaga-verify (crate iaga-sentinel-verify) is a separate, dependency-light binary. Export a run with iaga replay <run_id> --export run.json, then iaga-verify run.json --key <hex> checks the Ed25519 signatures and the Merkle chain offline, with no database and no IAGA binary. It is the artifact you hand an auditor. Build the slim verifier reproducibly with cargo build --release -p iaga-sentinel-verify --no-default-features --features verify-only.
---
iaga policy compile policy.apl --output policy.wasm
iaga plugins verify ./plugins/my-plugin.wasm
```bash cargo install --path crates/iaga-sentinel-core
docker compose up -d
curl http://localhost:4010/health # -> 200
docker compose down
The container persists its DB and signer key in a named volume (iaga-sentinel-data). Receipts signed inside the container can only be verified by the same container. To share a signer key across deployments, mount your own key file or set IAGA_SENTINEL_SIGNER_KEY_PATH.
```bash cargo install --path crates/iaga-sentinel-core
Verifiable by git clone && cargo test --workspace && docker compose up -d:
POST /v1/inspect), 275/275 default tests passing.iaga replay <run_id> --verify-only.iaga policy lint <file.apl>.iaga serve --policy <file.apl>. Stricter-wins merge with the YAML profile system.iaga reasoning info. Bring your own ONNX models via --features ml (tract backend, no native deps).UserspaceKernel, iaga run -- <cmd> spawns governed child processes on Linux, macOS, Windows.iaga gen-key then call POST /v1/inspect with Authorization: Bearer <key>.DATABASE_URL=postgres://... and building with --features postgres. Receipts go to the matching backend automatically.IAGA_SENTINEL_SIGNER_KEY_PATH points at any 32-byte Ed25519 key file, including one served by your KMS (AWS KMS, Azure Key Vault, HashiCorp Vault, on-prem HSM via the filesystem-mount pattern).docker compose up -d, /health returns 200 within about 10 seconds on the first attempt.iaga plugins list and iaga plugins validate <file.wasm>.Run the smoke yourself. Every claim above is reproducible from a clean checkout.
This walkthrough takes you from a clean checkout to a cryptographically signed, offline-verifiable record of an agent action, then layers on governance and observability. Every command and output below is real (captured from the open build on the default SQLite backend).
iaga serve --seed-demo ```
IAGA_SENTINEL_OPEN_MODE=true iaga serve --seed-demo
Build with --features ml, point IAGA_SENTINEL_REASONING_MODELS at your ONNX models, and the reasoning plane emits scores the policy can read. ML produces evidence, never the verdict; receipts embed the SHA-256 of every model that touched the decision.
```bash
```bash
iaga gen-key --label my-app
iaga run --agent-id openclaw-builder-01 -- python my_agent.py
Put IAGA Sentinel in the loop of any agent framework — one signed receipt per tool call. Adapters live in the SDKs (sdks/python, sdks/typescript) with copy-paste examples in examples/integrations/.
Shipped: Custom (@governed), LangChain, LangGraph (Py/JS), LlamaIndex, Pydantic AI, OpenAI Agents SDK, CrewAI, AutoGen, Microsoft Agent Framework, OpenAI (Py/TS), Vercel AI SDK, MCP (GovernedTool + iaga proxy), Claude Code (PreToolUse hook) and the Claude Agent SDK. Each is cooperative governance (allow / review / block, fail-open-by-default transport); a Rust client crate (iaga-sentinel-integrations) speaks the same wire contract.
The Python adapters are tested both with dependency-free fakes (CI) and against the real framework libraries (sdks/python/tests/e2e/). See the support matrix and per-framework guides in examples/integrations/README.md.
---
IAGA Sentinel has an open build and a commercial Enterprise edition. The open build is the source-verifiable evidence core in this repository. Enterprise adds managed, platform-specific, and compliance-delivery capabilities for organizations that need operational support beyond the public runtime.
The public boundary is documented in docs/adr/0010-oss-enterprise-boundary.md.
IAGA Sentinel 是 EU AI Act 合规性证据层,为 AI 代理的每个行动提供密码学签名、可离线验证的欧盟主权证明。所有操作记录都映射到 AI Act 第 12 条和附件 IV 的要求,确保代理行为的完整可追溯性和合规性。该项目通过 Merkle 链式收据和 Ed25519 签名机制,为 AI 系统提供企业级审计和治理能力。
核心功能包括:SQLite/Postgres 审计后端存储、Ed25519 签名的 Merkle 链式收据(M2)、Agent Policy Language(APL)策略解析与评估、WASM 编译支持、Sigstore 包验证、SBOM 集成、可选 ML 推理平面(ONNX 模型支持)。所有特性可通过 Cargo features 灵活启用,默认包含 SQLite、receipts 和 APL 功能。
需要 Rust 工具链用于源码编译。APL 编译为 WASM 模块需启用 `--features apl-wasm`。插件验证需启用 Sigstore 相关特性。Docker 部署需 Docker Compose。ML 推理功能需配置 ONNX 模型路径。系统需支持 Ed25519 密钥生成和密码学操作。
支持三种部署方式:(1)源码安装:`cargo install --path crates/iaga-sentinel-core`;(2)Docker Compose:`docker compose up -d`,容器通过命名卷持久化数据库和签名密钥;(3)自定义密钥:通过 `IAGA_SENTINEL_SIGNER_KEY_PATH` 环境变量挂载密钥文件实现跨部署共享。安装后可通过 `curl http://localhost:4010/health` 验证服务健康状态。
通过环境变量配置:`IAGA_SENTINEL_SIGNER_KEY_PATH` 指定签名密钥路径、`IAGA_SENTINEL_REASONING_MODELS` 指向 ONNX 模型目录(启用 ML 推理)。APL 策略文件通过 `iaga policy compile policy.apl --output policy.wasm` 编译。插件验证通过 `iaga plugins verify` 命令。支持 `--features ml` 启用机器学习推理平面,所有模型哈希值嵌入收据中。
提供 CLI 和 HTTP API 两种接口。CLI 无需认证,支持 `iaga run --agent-id <id> -- <command>` 启动治理管道。HTTP API 需通过 `iaga gen-key --label <name>` 生成 API 密钥。每个工具调用生成一份签名收据。支持 Python、TypeScript SDK 集成,兼容 LangChain、LangGraph、Llama 等主流代理框架。
工作流通过 `iaga run` 命令启动子进程并纳入治理管道,每次工具调用生成一份密码学签名的收据。支持多框架集成:Custom(`@governed` 装饰器)、LangChain、LangGraph(Python/JavaScript)、Llama 等。SDK 和集成示例位于 `sdks/` 和 `examples/integrations/` 目录,支持复制粘贴快速集成。所有证据链条可离线验证,满足 EU AI Act 审计要求。
IAGA Sentinel是一个开源的AI工作流工具,使用Rust语言开发,提供了AI代理无法回答的问题答案功能,提高AI工作流的可观察性和安全性。虽然其AI能力有限,但仍然是一个值得关注的项目。
该工具使用 NOASSERTION 协议,商用场景请仔细阅读协议条款,必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
📄 NOASSERTION — 请查阅原始协议条款了解具体使用限制。
经综合评估,IAGA-Sentinel 在Agent工作流赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | IAGA-Sentinel |
| 原始描述 | 开源AI工作流:IAGA Sentinel sits next to your AI agents and answers the one question the agent。⭐122 · Rust |
| Topics | workflowaiai-agentscybersecurityobservabilityrust |
| GitHub | https://github.com/EdoardoBambini/IAGA-Sentinel |
| License | NOASSERTION |
| 语言 | Rust |
收录时间:2026-06-09 · 更新时间:2026-06-11 · License:NOASSERTION · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端