AIR Blackbox

Install

pip install air-blackbox

Run the demo

Generates sample AI agent records and runs a compliance check. No Docker, no config.

air-blackbox demo

→ Creates 10 sample .air.json records (4 models, 3 providers)
→ Generates RISK_ASSESSMENT.md + DATA_GOVERNANCE.md templates
→ Runs EU AI Act compliance check: Articles 9-15

See your compliance status

air-blackbox comply -v

Shows per-article status with detection type (AUTO/HYBRID/MANUAL) and fix hints for every failing check.

Discover what's in your environment

air-blackbox discover

Shows every model, provider, and API endpoint your agents are using. Shadow AI detection built in.

Start the full gateway (optional)

For live traffic analysis with HMAC audit chains, tamper-proof logging, and real-time compliance:

git clone https://github.com/airblackbox/airblackbox.git
cd gateway
cp .env.example .env   # add your OPENAI_API_KEY
docker compose up

Then point your LLM client at the gateway:

from openai import OpenAI
from air_blackbox import AirBlackbox

air = AirBlackbox()
client = air.wrap(OpenAI())
# Every LLM call is now HMAC-logged through the gateway