AIR Blackbox
Open-source compliance infrastructure for AI agents
Tamper-proof audit trails · PII tokenization · Policy enforcement · AI Bill of Materials
AIR Blackbox is the compliance layer between your AI agents and production. It records every LLM call, tool invocation, and agent decision into tamper-evident audit chains — so you can prove what your AI did, when, and why.
Built for teams shipping AI agents that need to meet EU AI Act requirements by August 2026.
Your Agent Code
│
▼
┌──────────────────┐
│ AIR Gateway │ ← reverse proxy, records every LLM call
│ (Go binary) │
└────────┬─────────┘
│
┌────┴─────┬──────────────┬─────────────────┐
▼ ▼ ▼ ▼
Audit PII Policy AI Bill of
Chain Tokenization Engine Materials
(HMAC) (DataVault) (risk tiers) (CycloneDX)
│ │ │ │
└────┬─────┴──────────────┴─────────────────┘
▼
Compliance Report
(Articles 9–15)
pip install air-blackboxFramework trust layers (drop-in, no code changes):
pip install air-langchain-trust # LangChain / LangGraph
pip install air-crewai-trust # CrewAI
pip install air-openai-trust # OpenAI Agents SDK
pip install air-anthropic-trust # Claude Agent SDK
pip install air-adk-trust # Google ADK
pip install air-rag-trust # RAG pipelinesMCP server (Claude Desktop, Claude Code, Cursor):
pip install air-blackbox-mcpFull stack (Gateway + Episode Store + Policy Engine + Jaeger):
git clone https://github.com/airblackbox/air-platform
cd air-platform && make up| Repo | What it does |
|---|---|
| gateway | AI governance control plane — CLI scanner, reverse proxy, audit trails |
| air-platform | Full stack Docker Compose deployment (Gateway + Episode Store + Policy Engine + Jaeger) |
| air-gate | AI action firewall — HMAC-SHA256 audit chain with tool gating and PII redaction |
| air-blackbox-mcp | MCP server — 14 compliance tools for Claude Desktop, Claude Code, and Cursor |
| compliance-action | GitHub Action — EU AI Act compliance checks on every PR |
| Repo | Framework |
|---|---|
| air-langchain-trust | LangChain / LangGraph |
| trust-crewai | CrewAI |
| air-adk-trust | Google Agent Development Kit |
| air-rag-trust | RAG pipelines |
| Repo | What it does |
|---|---|
| otel-collector-genai | OTel Collector processor — PII redaction, cost metrics, loop detection |
| otel-prompt-vault | Encrypted prompt/completion storage with pre-signed URL retrieval |
| otel-semantic-normalizer | Normalizes gen_ai.* attributes to standard schema |
| runtime-aibom-emitter | AI Bill of Materials from runtime OTel traces (CycloneDX) |
| aibom-policy-engine | Policy-as-code for AI supply chain governance |
| mcp-security-scanner | Security scanner for MCP server configurations |
| mcp-policy-gateway | Policy enforcement gateway for Model Context Protocol |
| Repo | What it does |
|---|---|
| eval-harness | Replay and score agent episodes against policies |
| trace-regression-harness | Detect behavioral regressions across agent versions |
| Package | Version | Install |
|---|---|---|
| air-blackbox | 1.8.0 | pip install air-blackbox |
| air-gate | 0.2.0 | pip install air-gate |
| air-blackbox-mcp | 0.2.0 | pip install air-blackbox-mcp |
| air-blackbox-sdk | 0.1.1 | pip install air-blackbox-sdk |
| air-compliance | 1.0.0 | pip install air-compliance |
| air-langchain-trust | 0.2.0 | pip install air-langchain-trust |
| air-crewai-trust | 0.1.0 | pip install air-crewai-trust |
| air-openai-trust | 0.1.0 | pip install air-openai-trust |
| air-anthropic-trust | 0.1.0 | pip install air-anthropic-trust |
| air-adk-trust | 0.2.0 | pip install air-adk-trust |
| air-rag-trust | 0.1.0 | pip install air-rag-trust |
EU AI Act enforcement begins August 2, 2026. Start now.
