Get Started

OpenSearch AgentHealth helps you do two main things: Instrument your AI agents for observability, and Evaluate their performance with trajectory-based analysis.

Try the demo instantly:

npx @opensearch-project/agent-health@latest

# Open http://localhost:4001 in your browser

Deploy with OpenSearch backend:

git clone https://github.com/opensearch-project/dashboards-agent-health.git
cd dashboards-agent-health
docker compose up -d

# Access UI at http://localhost:4001

Add observability to your AI agents using OpenTelemetry. Capture traces, tool calls, and token usage.

Go beyond output comparison. Evaluate the entire agent trajectory with AI Judge scoring.

• Core Concepts → — Understand traces, trajectories, and benchmarks
• Example Project → — See a complete working example
• Integrations → — Connect with LangGraph, Claude Code, OpenAI, and more