Snippy

Detailed Introduction

Snippy is an open-source Azure Functions reference application that demonstrates exposing functions as MCP (Model Context Protocol) tools and coordinating stateful multi-agent workflows with Durable Agents. The project integrates Azure OpenAI for embeddings and LLM calls, Cosmos DB vector indexing for semantic code retrieval, and the Durable Task Scheduler for orchestrations. It provides labs and an azd workflow to provision and deploy the full stack locally or on Azure.

Main Features

• MCP tool integration: expose Azure Functions as discoverable tools for AI assistants.
• Multi-agent orchestration: coordinate specialized agents (e.g., DeepWiki, CodeStyle) using Durable Agents and DTS.
• Vector search: semantic code-snippet retrieval powered by Cosmos DB vector indexing.
• One-click deployment: azd up provisions functions, Cosmos DB, OpenAI and observability components.

Use Cases

• Learning & labs: hands-on tutorial to learn MCP tools, durable functions and vector search patterns.
• Developer assistants: demonstrate building tools discoverable by GitHub Copilot and similar assistants.
• Reproducible stacks: quick start templates for Codespaces, Dev Containers, and CI-driven deployments.

Technical Features

• Platform & language: Python-based Azure Functions with azd and Bicep infrastructure templates.
• Observability: DTS dashboard and telemetry for monitoring orchestration and tool invocations.
• Extensibility: modular tool matrix and pipeline design to swap model backends and retrieval layers.
• License: MIT, suitable for learning, reproduction and extension.