agentstate-sdk 0.1.0a1

Chrome DevTools for AI Agents — observability SDK for LangGraph, CrewAI, AutoGen

AgentStack

Chrome DevTools for AI Agents — open-source observability for LangGraph, CrewAI, AutoGen, and custom Python agents.

Install

pip install agentstate-sdk

Quick Start

from agentstack import observe, init

init(
    collector_url="https://your-collector.example.com",
    api_key="ak_your_project_key",
)

@observe
def research_agent(query: str) -> str:
    context = search_tool(query)
    return llm.chat(f"Answer based on: {context}")

@observe(name="planning.step")
async def async_agent(objective: str) -> list[str]:
    return await llm.achat(f"Break this into steps: {objective}")

Every call now produces a full trace — arguments, return values, timing, exceptions, token counts, and cost — visible in the AgentStack dashboard.

Key Features

• Real-time Tracing — Captures every LLM call, tool invocation, and function as structured spans
• Time Machine Replay — Step through past agent executions span-by-span
• Security Engine — Detects prompt injection, PII leakage, and anomalous behavior
• Cost Analytics — Per-model token counting with USD cost tracking
• Auto PII Sanitization — Scrubs sensitive data before export
• Framework Auto-Detection — Native hooks for LangGraph, CrewAI, and AutoGen
• Offline Resilience — Spans buffer locally when collector is unreachable
• Zero Interference — The @observe decorator never crashes your application

Supported Frameworks

Framework	Status
LangGraph	✅ Auto-instrumented
CrewAI	✅ Auto-instrumented
AutoGen	✅ Auto-instrumented
Custom Python	✅ Via @observe decorator

Self-Host the Dashboard

AgentStack is fully self-hostable with Docker Compose:

git clone https://github.com/Ramakrishna1967/AgentState.git
cd AgentState/deploy
cp .env.example .env
docker compose up -d --build

Links

• GitHub Repository
• Full Documentation
• Report Issues

License

Apache 2.0 — see LICENSE for details.