agentnotary
AgentNotary is an open-source CLI for governing AI agents across their lifecycle, providing cryptographic sealing, runtime enforcement, adversarial testing, and compliance documentation. It solves the problem of proving agent identity and preventing runaway costs or unsafe behavior, targeting developers building and deploying LLM-based agents.
✨ Key features
- Cryptographic seal (agent.lock) for agent snapshots
- Active proxy enforcement with guardrails (cost, PII, etc.)
- OWASP LLM Top 10 attack fuzzer
- Provider-drift detection via seal --probe
- Time-travel replay with fork and edit
- EU AI Act Annex IV compliance doc generation
🎯 Use cases
- Prevent runaway LLM costs with session cost caps
- Detect silent model updates by providers
- Generate audit-ready compliance docs for EU AI Act
- Fuzz agents for security vulnerabilities
- Enforce tool allowlists and PII redaction at runtime
📦 Installation
🧰 Requirements: Python 3.9+; works with any framework that respects ANTHROPIC_BASE_URL or OPENAI_BASE_URL; optional API keys for live LLM evals.
pip install agentnotarypip install "agentnotary[anthropic]" # for live LLM evals + attack runs
pip install "agentnotary[openai]"
pip install "agentnotary[pii]" # Presidio NER for stronger PII detection
pip install "agentnotary[all]"🚀 Usage
mkdir my-agent && cd my-agent
agentnotary init refund-botagentnotary doctor # health scan, score 0-100
agentnotary seal --probe # notarize + capture probe
agentnotary attack --suite owasp-llm-top10 # adversarial dry-run
agentnotary guard run -- python -m refund_bot # enforce at runtime
agentnotary compliance --standard eu-ai-act # certify
agentnotary bom --format cyclonedx # AI-BOM for procurement
git add agentnotary.yaml agent.lock docs/ && git commit -m "ship v0.1.0"⚠️ Good to know
Project is new (low stars) and some features like streaming proxy support are on the roadmap; no known limitations stated beyond that.
❓ FAQ
How does AgentNotary detect provider model drift?
It uses seal --probe to hash a canonical-prompt response, and drift re-probes to quantify any changes since the seal.
Can I use AgentNotary with any LLM framework?
Yes, it works with any framework that respects ANTHROPIC_BASE_URL or OPENAI_BASE_URL, including LangChain, CrewAI, AutoGen, and raw HTTP.
What does the guard command do?
It runs a local proxy that actively blocks runaway costs, off-allowlist tool calls, and PII leaks based on your guardrails configuration.
Is AgentNotary open source?
Yes, it is licensed under Apache 2.0 and can be used commercially, forked, or embedded.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.