Blog Post

Microsoft Foundry Blog
3 MIN READ

Foundry IQ: now managed for you in every Foundry project

fsunavala-msft's avatar
Nov 18, 2026

Foundry IQ: Now managed for you in every Foundry project

Building AI agents that can answer questions from your documents just got simpler.

Today we're announcing that every Microsoft Foundry project will include a managed Foundry IQ resource. Create knowledge bases and ground agents directly from your project, with no separate Foundry IQ (Azure AI Search) service to provision or manage.

With a managed Foundry IQ resource, it's easier to connect your agent to a knowledge base and activate market-leading agentic retrieval, ingestion, content security, indexing, and source connections within your Foundry agent-building experience. You choose the content; Foundry IQ manages the underlying grounding infrastructure.

Request access to the Foundry Managed Knowledge preview. We're starting with limited preview access and plan to roll out managed Foundry IQ to all Microsoft Foundry projects in December 2026.

How it works

Start with an eligible project and permissions to create a knowledge base and access the content you want to use.

Step 1: Open Knowledge in your Foundry project and select Create a knowledge base. In the configuration pane, use the Managed by Foundry resource and enter a name for your knowledge base.

 

Step 2: Select Upload files to add your files as a knowledge source. Foundry IQ processes the uploaded content and prepares it for retrieval through a searchable index. Once the knowledge source status changes to Active, save the knowledge base. You can also connect a knowledge source supported by the managed preview.

For supported file formats, upload limits, and processing details, see Create a file knowledge source.

 

 

Find evidence across documents with agentic retrieval

With model-based query planning enabled, Foundry IQ can break a complex question into focused searches, run them in parallel, and bring the relevant passages together for the agent.

For example, asking "How do our Q3 sales in Europe compare with our Q2 forecast?" may require retrieving actual sales results, forecast assumptions, and regional performance data from different documents. Foundry IQ can plan and execute focused searches across those sources to gather the evidence needed for a grounded response.

Source references let users check where the answer came from. Learn how agentic retrieval works.

Use microsoft-retriever-1 out of the box for fast, low-cost retrieval, or bring your own supported model to tune the trade-off between latency, cost, and answer quality. Different workloads may benefit from different models depending on query complexity, retrieval requirements, and performance goals.

Ground your agent and test its answers

From the saved knowledge base, select Add to an agent.

In the agent playground, ask a question that draws on multiple files and open a citation to check that the source supports the answer.

Then try a question answered by one document and another the documents can't answer. For the latter, check that the agent acknowledges the gap.

 

 

Choose how to run Foundry IQ

Use a managed Foundry IQ resource when you want to create a knowledge base in your project and have Foundry manage the search infrastructure.

Bring your own Foundry IQ (Azure AI Search) resource when you want to reuse existing search indexes or indexing pipelines, define custom index schemas, or control the search resource configuration directly.

Supported knowledge sources and configuration options vary by setup.

Get started

Request access to private preview and try grounding an agent with a set of documents you know well.

Additional resources:

Updated Nov 18, 2026
Version 1.0
No CommentsBe the first to comment