Features: Foundry Local

Scope

This is the features page for Foundry Local, one of the three targets in ADR-0011. Compare all three targets on the Deployment targets hub.

This target has not been built yet

Nothing on this target has been deployed from this repository, and no automation exists for it yet. Everything below is drawn from accepted decisions and first-party research. Treat it as a design, not an as-built record.

What is decided today

• An OpenAI-compatible local API, plus the foundry CLI and the Foundry Local SDKs.
• Whisper speech to text is available. Image generation and text to speech are not.
• Runs fully disconnected once the model cache is populated.
• No quotas and no rate limits. Capacity is bounded by the host.
• No content filter or responsible-AI guardrail is documented at all. This is the governance-relevant gap: a workload moved here from Azure AI Foundry silently loses every control ADR-0007 relies on.
• No Foundry Agent Service and no MCP surface. Tool calling works at the API level, so a bring-your-own-runtime agent loop can orchestrate its own tools.
• No fine-tuning service and no batch API. Microsoft states plainly that continuous batching is a capability Foundry Local does not provide. Olive can convert and optimize a model, and a model fine-tuned elsewhere can be brought here (SPIKE-31).
• No vision input and no embeddings. Structured outputs (json_schema) are not documented on chat completions.

The research and decisions behind this page

These were open when this page was first written. They are now written and linked below. Where a finding contradicted an earlier claim on this page, the page has been corrected and says so inline.

• SPIKE-31, the cross-track feature parity spike