Foundry Third-Party Agents Samples

Reference implementations that show how to plug Azure Application Insights into agents running on AWS Bedrock, Google Cloud, and the A2A protocol. Every sample has an end-to-end README plus an .env.example you can copy to start experimenting locally.

What’s inside

• aws/agent_core/ – LangGraph + AgentCore currency agent that calls Bedrock models and, when available, streams traces to Azure Application Insights.
• gcp/cloud_run_agent/ – FastAPI service that hosts a multi-stage LangGraph travel planner on Cloud Run with Gemini, optional Azure tracing, and a helper deploy.py.
• gcp/vertex/ – Vertex AI Agent Engines sample that builds a LangChain LangchainAgent, wires in a currency-rate tool, and forwards spans to Azure when configured.
• gcp/a2a_agent/ – Remote LangGraph currency agent packaged with the A2A server runtime, Dockerfile, deployment helper, and its own README.
• evals/ – Evaluation utilities, including trace/trace_eval.py for Application Insights traces and a2a/a2a_agent_eval.py for driving the remote agent over A2A connections.

How to use the repo

1. Pick the scenario you care about and open its README for architecture, setup, and deployment notes.
2. Copy the matching .env.example file, fill in your cloud credentials, and install the requirements listed for that sample.
3. Follow the per-sample README to run locally or deploy; when APPLICATION_INSIGHTS_CONNECTION_STRING (or equivalent) is set, telemetry flows into Azure automatically.