Metadata-Version: 2.4
Name: agentaudits
Version: 0.1.0
Summary: Cryptographic receipts for every AI agent action. Hash-chained, tamper-evident, regulator-ready.
Project-URL: Homepage, https://agentaudit.co.uk
Project-URL: Documentation, https://agentaudit.co.uk/docs
Project-URL: Issues, https://agentaudit.co.uk/contact/
Project-URL: Source, https://github.com/Kymeira-MrH/agent-audit
Author-email: VantagePoint Networks <info@vpnetworks.co.uk>
License-Expression: LicenseRef-Proprietary
License-File: LICENSE
Keywords: agent,ai,anthropic,audit,claude,compliance,eu-ai-act,mcp,observability,openai
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: Other/Proprietary License
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Security :: Cryptography
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: >=3.11
Requires-Dist: cryptography>=42.0.0
Requires-Dist: httpx>=0.27.0
Requires-Dist: pydantic>=2.6.0
Provides-Extra: dev
Requires-Dist: jsonschema>=4.21.0; extra == 'dev'
Requires-Dist: mypy>=1.10.0; extra == 'dev'
Requires-Dist: pytest-asyncio>=0.23.0; extra == 'dev'
Requires-Dist: pytest-cov>=5.0.0; extra == 'dev'
Requires-Dist: pytest>=8.0.0; extra == 'dev'
Requires-Dist: ruff>=0.5.0; extra == 'dev'
Requires-Dist: types-jsonschema>=4.21.0; extra == 'dev'
Description-Content-Type: text/markdown

# agentaudit

> Cryptographic receipts for every AI agent action.
> Hash-chained, tamper-evident, regulator-ready.

Drop-in Python SDK for capturing audit-grade evidence of AI agent activity.
Works with OpenAI Agents SDK, Claude Agent SDK, MCP, LangChain, CrewAI, or
any custom agent runtime.

**Status:** Alpha. Pre-launch. Open beta Q3 2026.

## Why

Your existing logs were not built for AI agents.

- Datadog and SIEM capture HTTP calls. They do not capture which tool an
  agent invoked, which decision it made, or what data classification it
  touched.
- The EU AI Act Article 12 logging obligations land August 2026 — and
  your current logs are unlikely to satisfy a regulator.
- Cyber and E&O insurers are already asking about AI logging at renewal.

`agentaudit` produces tamper-evident receipts of every agent action, with
cryptographic integrity guarantees, that satisfy regulators, insurers,
and auditors.

## Install

```bash
pip install agentaudits
```

You will need an Agent Audit account and an API key. Sign up at
<https://www.agentaudit.co.uk/signup/> — the Free Dev tier covers
100,000 receipts/month and is enough for evaluation.

## Quick start

```python
import agentaudit

agentaudit.init(
    api_key="aa_live_...",
    endpoint="https://api.agentaudit.co.uk",
    agent_id="claims_triage_v3",
)

# Your agent runs as normal. Receipts are captured automatically.
```

For framework-specific adapters (OpenAI Agents SDK, Claude Agent SDK, MCP),
see [agentaudit.co.uk/docs/sdk-py](https://agentaudit.co.uk/docs/sdk-py).

## Properties

Every receipt has three integrity guarantees:

1. **Inline PII redaction.** Raw PII never leaves your perimeter.
2. **Hash-chained.** Each receipt embeds the previous receipt's hash;
   retroactive tampering is mathematically detectable.
3. **Optionally notarised.** The chain head can be signed hourly via
   RFC 3161 timestamping for court-admissible cryptographic proof.

## Licence

**This SDK is proprietary commercial software.** Licensed, not sold, under
the Agent Audit Commercial Licence Agreement — see [LICENSE](LICENSE).

Notable restrictions:

- The SDK may be used **only** by holders of an active Agent Audit
  account (including the Free Dev tier) and **only** to send receipts to
  that account.
- **No reverse-engineering**, decompilation, derivation, or
  reconstruction of source code, file format or algorithms.
- **No use** of the SDK or any output of the Service to develop, train,
  evaluate or operate a competing product.
- **No redistribution**. The SDK may not be hosted, shared, sublicensed,
  resold or made available to any third party.

Source-level access to the SDK is available **only to paid Customers**
under a separate written non-disclosure agreement, for the limited
purpose of internal security review by their engineering and information-
security teams.

The receipt schema (`schemas/receipt.v1.json`) and the verification
protocol are intentionally published so that auditors, regulators and
customer DPOs can independently verify the integrity of receipts. This
public schema does **not** grant any rights in the SDK or the backend.

Full Terms of Service:
<https://www.agentaudit.co.uk/legal/terms/>

## Security

For security issues, contact `info@vpnetworks.co.uk`. Please do not file
public issues for vulnerability reports.

— VantagePoint Networks, United Kingdom
