Microsoft used their Build 2026 event in San Francisco to announce new functionality for Microsoft Foundry. Citing Foundry as "the place where AI agents move from experiments to production systems," in a blog post, Nick Brady writes that the release brings 'runtime, tools, memory, grounding, models, observability, and governance' that developers need for production agents, rather than just new model endpoints.
Microsoft Foundry is Microsoft’s 'AI app and agent factory', a unified Azure platform that Microsoft describes as an interoperable platform that lets teams build, ground, and govern AI apps and agents that understand business context, with shared observability and policy across every agent. Foundry documentation highlights native integration with Azure services, Microsoft 365 data sources, and open protocols for tools and frameworks.
Hosted agents in Foundry Agent Service provide managed sandboxed sessions with state, filesystem access, and support for multiple frameworks, exposing both a stateful Responses API and a lighter weight invocations protocol for passthrough calls. The same runtime supports long-running agents such as OpenClaw and Hermes with durable state and files, and routines, currently in public preview, to run agents on schedules for tasks like overnight ticket triage or daily reporting. These additions extend the Azure AI Foundry Agent Service general availability release that InfoQ covered in 2025, which introduced multi agent orchestration, Agent to Agent APIs, and support for frameworks such as Semantic Kernel, AutoGen, and CrewAI.
The blog turns to speak of tooling and distribution. Toolboxes in Foundry, now in public preview, give agents a single managed endpoint for tools, skills, Model Context Protocol clients, and enterprise data integrations, so that tools are registered once and then discovered at runtime rather than wired into each agent. Skills can be versioned, project-scoped artefacts cab be exposed via MCP, and tool search helps the platform select a small set of relevant tools for each task instead of exposing everything to the model. Microsoft is also introducing direct publishing from Foundry into Microsoft Teams and Microsoft 365 Copilot, with general availability planned for June 2026, so that agents built in Foundry can be surfaced where employees already work with identity, permissions, and policy applied automatically.
Foundry treats "memory" as a function of the platform, rather than just in applications. Memory in Foundry Agent Service, which entered public preview at the end of 2025, now offers procedural, user, and session memory. Procedural memory, new at Build, is designed to help agents learn how to carry out work across runs, with early benchmarks showing improvements in task success rates when it is enabled. InfoQ’s earlier coverage of Foundry memory describes how the service extracts key facts and procedures from conversations, consolidates them, and then retrieves them through a managed store scoped by identifiers such as Entra ID, with controls for retention and inspection.
Procedural memory helps agents learn how to do the work across runs, not just what was said, with early Tau bench results showing 7 to 14 percent absolute success rate gains at near baseline cost.
--Nick Brady
Grounding and retrieval are tackled through Foundry IQ, which Brady presents as a knowledge layer behind agents that unifies Work IQ, Fabric IQ, Azure SQL, file search, and other sources behind a single SLA backed retrieval endpoint. At Build, Microsoft announced Foundry IQ Serverless in public preview, multi source knowledge bases in general availability, and Microsoft Web IQ for live web grounding with sub 200 millisecond responses and zero data retention guarantees, together with security features around encryption, permission synchronisation, and sensitivity label governance. In a separate deep dive, Satyanarayana Padidapu describes Microsoft IQ, which covers Work IQ, Fabric IQ, and Foundry IQ, as an 'intelligence layer' that aims to reduce duplicated retrieval augmented generation pipelines and make grounding a shared service across Copilot Studio, Microsoft 365, and Foundry agents.
On the model side, Foundry’s catalogue adds four first party MAI models in public preview: MAI Thinking 1 for chat and reasoning, MAI Image 2.5 for image generation and editing, MAI Transcribe 2 for speech to text with diarisation, and MAI Voice 2 for multilingual text to speech with voice cloning. Fireworks AI on Foundry is now generally available, providing access to open models through a single Azure endpoint with enterprise SLAs, support for custom weight models, and integration with Foundry’s access controls and logging. Vesa Nopanen’s analysis of Claude Opus on Foundry calls this pattern 'a meaningful step forward' for organisations that want frontier models with Azure governance, highlighting day zero access, low latency, and integration with Foundry IQ and Work IQ for grounded agents. Managed Compute in Foundry Models is aimed at routing workloads around regional GPU constraints and supports fine tuning and 'Frontier Tuning', which Microsoft claims is significantly more cost efficient than using GPT 5.5 directly for tasks such as technical documentation generation.
Tracing and evaluations for any agent framework mean no team has to choose between its stack and observability. You can keep LangChain, Semantic Kernel, or your own code, and still get production grade traces and evaluations in Foundry.
--Nick Brady
Outside the Build recap, Microsoft and community authors describe a layered picture in which Microsoft 365 Copilot Agent Builder and Copilot Studio provide visual, low code experiences, while Foundry is the code first platform with evaluations and observability. Szymon Bochniak’s comparison of Agent Builder, Copilot Studio, and Foundry presents this as three tiers, with Foundry used when teams need custom logic, advanced retrieval, and deeper integration with developer workflows. Microsoft's secure agent process guidance recommends that teams map where agents already touch build, test, and release, and then apply the same discipline they use for microservices: clear scope, policy, tracing, and continuous evaluation, now supported as first class features in Foundry.And a Build 2026 recap written from a DevOps perspectivedescribes these additions as the point where "Foundry feels like a real production platform for agents, not just a place to wire up demos."
Further information on Foundry is available on Microsoft's website.