The Security Firewall for LLM Applications

All endpoints require an API key in the request header:

X-API-Key: YOUR_API_KEY

Roles: standard → analyst → admin

In Swagger UI: click Authorize (top right), paste your key, click Authorize then Close.

1. Get your API key (free plan above)

2. Inspect a prompt:

POST /inspect/prompt
X-API-Key: YOUR_KEY

{"prompt": "your user input", "agent_id": "app"}

3. Check decision in the response — allowed or blocked

4. Pass the session_id to POST /inspect/response to also scan the LLM output.

import httpx

SENTINEL = "https://sentinelai.expert"
HEADERS  = {"X-API-Key": "YOUR_KEY",
            "Content-Type": "application/json"}

def safe_llm_call(prompt: str) -> str:
    check = httpx.post(
        f"{SENTINEL}/inspect/prompt",
        headers=HEADERS,
        json={"prompt": prompt, "agent_id": "app"}
    ).json()
    if check["decision"] == "blocked":
        return "Blocked: " + str(check.get("reasons"))
    return your_llm(prompt)

Connect to /ws/feed to receive live threat events:

wss://sentinelai.expert/ws/feed
Header: X-API-Key: YOUR_KEY

Events fire every time a threat is detected. Ping/pong keepalive every 15 seconds. Max 100 simultaneous connections.

Create: POST /keys — returns raw key once only

List: GET /keys — metadata only, raw values never exposed

Revoke: DELETE /keys/{"{key_id}"} — revoked keys cannot authenticate

⚠️ Never use your admin key in application code. Create a dedicated standard key per app.