Summary This repository hosts a deprecated, experimental Model Context Protocol (MCP) server for Azure AI Foundry, offering a unified interface for AI tool orchestration. It targets developers and agents needing streamlined interaction with models, knowledge, and evaluation services within Azure AI Foundry, aiming to accelerate productivity via natural language workflows. This specific implementation is no longer maintained.

How It Works The MCP Server acts as a cloud-hosted endpoint, exposing tools for managing Azure AI Foundry resources. It adheres to the Model Context Protocol, abstracting backend complexities. This allows clients, including AI agents, to discover and invoke operations for model exploration, knowledge indexing (Azure AI Search), evaluation, and fine-tuning without direct API knowledge, facilitating integration and multi-agent scenarios.

Quick Start & Requirements

• Installation: Requires uvx (installable via uv). A GitHub template is provided for minimal setup with MCP server configuration and dependencies.
• Configuration: Setup involves creating a .vscode/mcp.json file in VS Code to define the server command (uvx run-azure-ai-foundry-mcp) and arguments, referencing an optional .env file for environment variables. Advanced setup details are available.
• Prerequisites: uvx installed; VS Code recommended.
• Environment Variables: Key variables include GITHUB_TOKEN (optional, for model testing), AZURE_AI_SEARCH_ENDPOINT and auth details for knowledge indexing, and Azure OpenAI credentials for text evaluation.
• Links: README directs users to the new Foundry MCP Server (preview) documentation: Get started with Foundry MCP Server (preview).

Highlighted Details

• Comprehensive Tooling: Exposes capabilities for exploring models, managing Azure AI Search indexes (CRUD operations on indexes, documents, data sources, indexers), running text/agent evaluations, and monitoring fine-tuning jobs.
• Secure Orchestration: Integrates Microsoft Entra ID for authentication and authorization via On-Behalf-Of flow.
• Developer Productivity: Accelerates workflows via natural language tool discovery and invocation, simplifying interaction with AI services.

Maintenance & Community This repository contains outdated, experimental code and will not be updated. All development has transitioned to the cloud-hosted "Foundry MCP Server (preview)". No community links or contributor details are provided for this deprecated repository.

Licensing & Compatibility

• License: MIT License.
• Compatibility: MIT license generally permits commercial use and integration into closed-source projects, but the project itself is deprecated.

Limitations & Caveats This repository is explicitly marked as outdated and experimental, receiving no further updates. It serves solely as a historical reference. Users should refer to the new, remotely hosted "Foundry MCP Server (preview)" for current functionality and support.