mcp-foundry
mcp-foundry is an MCP (Model Context Protocol) server (Python/stdio) intended to provide a unified set of tools for Azure AI Foundry, covering model discovery/details, deployment-related operations, knowledge/index management for Azure AI Search, evaluation utilities for text/agents, and fine-tuning/job status and related management. The README indicates this repo is a deprecated/older experimental implementation and that a newer Foundry MCP Server exists as a cloud-hosted preview.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
The README advises using a .env file for sensitive values (suggesting secret management via environment variables rather than hard-coding). However, it does not provide strong evidence of how secrets are logged/handled internally, nor does it describe fine-grained authorization boundaries/scopes in this local server. Dependency list includes several Azure libraries; without vulnerability evidence, hygiene is inferred as moderate.
⚡ Reliability
Best When
You want agent-driven orchestration of Azure AI Foundry and related Azure services from an MCP client, and you can supply the required environment variables for the targeted operations.
Avoid When
You require high operational stability/long-term support for this specific repository, or you need an officially supported cloud-hosted Foundry MCP endpoint rather than a local/deprecated implementation.
Use Cases
- • Let an MCP client (e.g., Copilot chat agent mode) discover and invoke Azure AI Foundry-related tools for models, deployments, evaluations, and agents
- • Automate Azure AI Search index/schema/indexer/data source/skill set management from an agent workflow
- • Run evaluation workflows (text evaluators, agent evaluation, formatting reports) using MCP tools
- • Monitor and manage Azure fine-tuning jobs (status, events, metrics, files) via agent-invoked tools
Not For
- • Production systems that require a stable, maintained API surface (README states the repository will not be updated further)
- • Teams that cannot provide required Azure credentials/endpoints or prefer not to handle secrets locally
- • Use cases needing a web-hosted, externally reachable endpoint (this is configured as an MCP stdio server run locally)
Interface
Authentication
The README shows configuration via environment variables. For Azure AI Search it supports 'service-principal' (tenant/client/secret) or 'api-search-key'. The deprecated local MCP server does not advertise Microsoft Entra on-behalf-of; it appears to rely on the supplied credentials. The README also mentions a separate cloud-hosted Foundry MCP Server that enforces Microsoft Entra ID.
Pricing
README mentions GitHub token for free testing of models with rate limits, but no quantified limits or pricing tiers are provided.
Agent Metadata
Known Gotchas
- ⚠ Repository is marked experimental/deprecated; it may stop receiving updates in favor of the cloud-hosted Foundry MCP Server
- ⚠ Many operations require correctly set environment variables; missing/incorrect credentials can cause tool failures
- ⚠ State-changing actions (create/delete/modify) can be risky in agent loops because idempotency and retry guidance are not documented
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for mcp-foundry.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.