Azure AI Foundry architecture

A governed gateway for every agent call

Azure AI Foundry runs agents and models. Kimss adds the workspace control layer around that execution: identity, tenant routing, usage attribution, and Kimss Credits.

Execution plane and control plane

Route by workspace

Map authenticated users and API keys to the correct Foundry project or shared shard.

Meter consistently

Normalize model usage into Kimss Credits and attribute it to tenants, agents, and billing pools.

Enforce boundaries

Keep workspace identity and permissions attached to calls instead of relying on a single blended Azure resource.

One preferred developer surface

The universal gateway uses routes such as /v1/agents/run and /v1/models/completions. Clients authenticate with X-Kimss-Key or a bearer token and can identify a workspace with X-Workspace-ID.

Azure API Management is an optional path in the Kimss architecture. Direct Foundry routing remains the safe default until an environment’s APIM path and telemetry have passed end-to-end verification.