v1.3.0 · MIT License · Stable Beta
Runtime governance
for AI agents.
Zentinelle sits between your AI agents & their LLM providers
enforcing policies
logging every action
real-time compliance dashboard
Architecture
In the middle of everything.
Invisible to your agents.
Zentinelle intercepts every LLM call and tool use — applying your policies in real time, logging results, and reporting compliance. Your agents keep running unchanged.
Policies enforced:rate limitstool permissionsPII detectionmodel restrictionscost controls
Your Agent
Claude CodeCodexGemini CLICustom
Zentinelleintercepting
✦Policy Check
≡Audit Log
◎Monitoring
LLM Provider
AnthropicOpenAIGoogle
Capabilities
What Zentinelle gives you.
Policy enforcement, observability, compliance, and enterprise resilience — all in one runtime layer.
Policy Enforcement
Rate limits, cost controls, tool allowlists/blocklists, model restrictions, PII detection and masking, human-in-the-loop approvals.
Real-time Observability
Every tool call, every LLM invocation — visible in the dashboard. Token usage, latency, error rates, and agent behavior in one place.
Compliance Frameworks
Built-in dashboards for SOC2, GDPR, HIPAA, EU AI Act, and NIST AI RMF. Audit trails that satisfy reviewers without manual work.
Audit Logging
Immutable logs of every agent action, policy decision, and LLM response. Know exactly what your agents did and why it was allowed.
Enterprise Resilience
Circuit breaker pattern, retry with exponential backoff, fail-open mode for non-critical paths, event buffering and batching.
Multi-agent Support
Track Claude Code, Codex, Gemini CLI, and custom agents side-by-side. One governance layer for your entire AI agent fleet.
Project Status
v1.0. Production ready.
LIVE
Claude Code
LIVE
OpenAI Codex
LIVE
Gemini CLI
LangChain
LIVE
LlamaIndex
Soon
CrewAI
LIVE
Vercel AI SDK
LIVE
n8n
Soon
Microsoft Agent Framework
Soon
Zentinelle v1.0 is here. 24 policy evaluators, 5 SDK languages, 3 framework plugins, multi-cloud Terraform. MIT licensed, self-hostable, production-ready.
Working now1 / 8
Claude Code integration (hooks)
Roadmap
Coming soon.
Built in the open.
LangChain
LIVE
LlamaIndex
Soon
CrewAI
LIVE
Vercel AI SDK
LIVE
n8n
Soon
Microsoft Agent Framework
Soon
Everything on this list is being built in public. Star the repo, open issues, and help us prioritise what matters most to your team.
Coming soon1 / 9
LangChain callback handlers
Quick Start
Up in five minutes.
Clone the repo, run Docker Compose, point your agent at the proxy. That's it.
Clone and start
# 1. Clone and start
git clone https://github.com/calliopeai/zentinelle
cd zentinelle
docker compose up -d
# 2. Initialize (first run only)
docker compose exec backend python manage.py migrate
docker compose exec backend python manage.py createsuperuser
# 3. Open the dashboard
open http://localhost:8080
Connect your agent (proxy mode — works with any agent)
# 4. Connect your agent (proxy mode — works with any agent)
export ANTHROPIC_BASE_URL=http://localhost:8080/proxy/anthropic/
export OPENAI_BASE_URL=http://localhost:8080/proxy/openai/
# Every LLM call is now governed. Dashboard: http://localhost:8080