🔭 AgentixLens

AI Agent Observability — trace, monitor, and debug your AI agents in real-time.

🌐 Website · 📊 Dashboard · 🐦 Twitter

---

What is AgentixLens?

AgentixLens gives you a transparent window into every decision your AI agents make.

• Visual Trace Explorer — every LLM call, tool use, and decision branch in a waterfall
• Cost & Token Tracking — per-run, per-model, per-tool granularity
• Latency Heatmaps — p50/p95/p99 across your entire agent fleet
• Failure Capture & Replay — reproduce any failed run with full context
• Model-Agnostic — works with any LLM, zero vendor lock-in

---

Repo Structure

agentixlens/
├── index.html          # Landing page (deployed to Netlify)
├── dashboard.html      # Dashboard UI (deployed to Netlify)
├── netlify.toml        # Netlify config
├── backend/            # FastAPI backend (deploy to Railway/Render/Fly)
│   ├── main.py
│   ├── requirements.txt
│   ├── Dockerfile
│   ├── .env.example    # Copy to .env and fill in values
│   ├── db/
│   ├── models/
│   ├── routers/
│   └── middleware/
└── sdk/                # Python SDK (publish to PyPI)
    ├── agentixlens/
    ├── examples/
    └── tests/

---

Quick Start

1. Install the SDK

pip install agentixlens

2. Instrument your agent

from agentixlens import lens, trace

lens.init(project="my-agent")

@trace("research-agent")
async def run_agent(query: str) -> str:
    # Your existing agent code — unchanged
    ...

3. Run the backend locally

cd backend
pip install -r requirements.txt
cp .env.example .env        # Fill in your values
uvicorn main:app --reload --port 4317

4. Open the dashboard

Open dashboard.html in your browser — it connects to http://localhost:4317 automatically.

---

Deployment

Frontend (Netlify) — free

1. Push this repo to GitHub
2. Connect to Netlify → Import from Git
3. Set Publish directory to . (root)
4. Set Build command to (empty)
5. Deploy — index.html and dashboard.html go live instantly

Backend (Railway) — one click

Or manually:

cd backend
railway login
railway up

Set these environment variables on Railway:

ENV=production
AUTH_ENABLED=true
API_SECRET_KEY=<generate with: python -c "import secrets; print(secrets.token_urlsafe(32))">
DASHBOARD_TOKEN=<generate same way>
ALLOWED_ORIGINS=https://agentixlens.com

Backend (Docker)

cd backend
docker build -t agentixlens-backend .
docker run -p 4317:4317 \
  -e API_SECRET_KEY=your-key \
  -e DASHBOARD_TOKEN=your-token \
  -v agentixlens-data:/data \
  agentixlens-backend

---

SDK Reference

from agentixlens import lens, trace, trace_llm, trace_tool, current_span

# Initialize
lens.init(
    project="my-agent",
    endpoint="https://your-backend.railway.app",
    api_key="ax_...",
    debug=True,
)

# Trace entire agent run
@trace("my-agent", tags={"env": "prod"})
async def run_agent(query: str): ...

# Trace individual LLM call
@trace_llm(model="claude-3-5-sonnet", provider="anthropic")
async def call_llm(messages): ...

# Trace tool calls
@trace_tool("web_search")
async def search(query: str): ...

# Add metadata from inside any function
current_span().set_attribute("user_id", user.id)

LangChain Integration

from agentixlens.integrations.langchain import AgentixLensCallback

llm = ChatOpenAI(callbacks=[AgentixLensCallback()])

---

Environment Variables

See backend/.env.example for all options.

Variable	Required	Description
API_SECRET_KEY	Yes (prod)	SDK authentication key
DASHBOARD_TOKEN	Yes (prod)	Dashboard UI access token
ENV	No	development or production
AGENTIXLENS_DB	No	SQLite path (default: ~/.agentixlens/server.db)
ALLOWED_ORIGINS	No	Comma-separated CORS origins
AUTH_ENABLED	No	Set false for local dev

---

Contributing

1. Fork the repo
2. Create a feature branch: git checkout -b feature/my-feature
3. Run tests: cd sdk && pytest tests/ -v
4. Commit and push
5. Open a Pull Request

---

License

MIT © 2025 AgentixLens

---

Built for developers who are tired of flying blind with AI agents.