Microsoft Foundry IQ Build Smarter Agents Faster with Serverless Knowledge and MCP

Why Foundry IQ Matters for AI Agent Development

Building production-ready AI agents is harder than it looks. The agent logic might be solid, but the knowledge layer underneath—the data, retrieval, security, and scaling—often becomes the bottleneck. Microsoft's answer is Foundry IQ, a managed knowledge platform that grounds agents in your company's collective intelligence: documents, emails, meetings, operational data, and even the live web.

At Microsoft Build 2026, Foundry IQ announced several major updates designed to speed up agent development and reduce infrastructure overhead. The key theme: provision knowledge bases faster, unify enterprise and external sources, and expose everything via the Model Context Protocol (MCP) so any agent framework can tap in.

If you're evaluating how to move from a prototype to a production agent, this update is worth a close look. Let's break down what's new and what it means for your stack.

What's New in Foundry IQ: Serverless, New Sources, and MCP

Foundry IQ Serverless (Public Preview)

Agent workloads are bursty by nature—hundreds of steps in seconds, then idle for hours. Serverless eliminates the need to manage clusters or reserve capacity. You only pay for compute and storage used, and the service scales to zero when idle.

Pricing snapshot (public preview):

Billing begins September 13, 2026. Use the Azure portal's Scale + Cost tab to estimate usage.

New Knowledge Sources (Preview)

• Work IQ: emails, meetings, files, Teams messages—respecting user permissions.
• Fabric IQ: query data agents and ontologies for structured answers.
• File Search: upload files directly.
• Azure SQL: bring relational data in.
• MCP Server: connect any MCP-compatible host.

Knowledge Bases Now Generally Available

Production-ready with SLA, compliance certifications, and stable APIs. The Foundry IQ MCP server works with Claude, ChatGPT, LangChain, and the Microsoft Agent Framework.

Agentic Retrieval Improvements

Answer quality improved by up to 20% across datasets, with 54% better recall than single-shot RAG—while using fewer tokens thanks to server-side caching.

Security, Data Pipeline, and What to Watch Out For

Security Updates (Preview)

• Cross-tenant customer-managed keys (CMK) with federated identity
• Purview sensitivity-label auditing
• Incremental SharePoint permissions sync
• APIM support for Foundry model integrations

Data Pipeline Updates (Preview)

• SharePoint indexing for ASPX pages and Lists
• Layout-aware ingestion via Azure Content Understanding
• Image verbalization and enrichment

Limitations & Cautions

• Serverless is still preview: pricing and features may change before GA.
• Region availability: not all regions offer serverless yet.
• MCP ecosystem maturity: while MCP is promising, not all hosts fully support remote MCP servers yet.
• Data governance: even with CMK, you must audit label propagation end-to-end.

For a deeper dive on security best practices, check our Critical React Server Components RCE Vulnerability Guide to see how we approach vulnerability response.

Conclusion: Start Small, Scale Fast

Foundry IQ brings a compelling value proposition: build once, reuse everywhere. With serverless, you avoid idle costs; with MCP, you avoid vendor lock-in; with unified sources, you avoid custom connectors.

Next steps for your team:

1. Create a Foundry IQ resource in the Foundry portal (serverless tier).
2. Connect your most critical knowledge source (e.g., SharePoint or Azure SQL) and test retrieval quality.
3. Experiment with the MCP server using a simple agent in LangChain or ChatGPT.
4. Monitor token consumption and answer quality—then scale to production.

If you're planning infrastructure migration, our Azure Storage Migration Strategy Guide offers a complementary perspective on moving data safely.

Source: Microsoft Foundry Blog