Security-first LLM gateway and guardrails framework
Put one protected gateway in front of your AI traffic
SentinelGuard helps teams secure LLM applications, AI IDEs, agents, and model provider traffic with prompt scanning, output scanning, PII and secret protection, model routing, failover, usage controls, audit events, and a stable management API.
Install and start the gateway
pip install "sentinelguard[gateway,monitoring]"
export OPENAI_API_KEY="sk-..."
export SENTINELGUARD_GATEWAY_API_KEY="$(sentinelguard token)"
sentinelguard init
sentinelguard gateway \
--config sentinelguard.yaml \
--gateway-config sentinelguard-gateway.yaml \
--port 8080Point apps and IDEs to http://localhost:8080/v1
Why teams use SentinelGuard
Runtime protection for the LLM boundary
Secure prompts and responses
Scan prompts before they reach a model and scan responses before they return to users. Block attacks, redact PII, and stop secrets from leaving the application.
Route across providers
Use friendly model names, route traffic across OpenAI-compatible providers, fail over when one provider is unavailable, and keep private routes available for sensitive traffic.
Operate with evidence
Use virtual keys, usage accounting, provider health, privacy-safe
audit events, Prometheus metrics, and the stable
/gateway/v1 API for dashboards and alerts.
Built for modern AI applications
Use it in code or as a shared gateway
Package mode
Use SentinelGuard directly in Python code when you want guardrails inside one application.
from sentinelguard import SentinelGuard
guard = SentinelGuard()
result = guard.scan_prompt("Ignore previous instructions")
print(result.is_valid, result.failed_scanners)Gateway mode
Run SentinelGuard as a proxy so multiple applications, users, and AI tools share one security boundary.
App or IDE -> SentinelGuard /v1 -> LLM providerStable endpoints: /gateway/v1/contract, /gateway/v1/health, /gateway/v1/routes
Explore the docs