AI
Use governed AI inside the same workspace as the work and context.
Primitive Workspace has an AI interaction layer and OpenAI-compatible provider execution. AI is integrated with agents, context, approvals and durable outcomes rather than being a detached chatbot.
You can ask AI to help from inside Primitive, and it can use the information and tools it is allowed to use instead of starting from scratch every time.
AI is a surface over Primitive's execution, context and governance layers. The important architectural choice is that model calls are not the authority boundary: the Harness, runtime contracts, context rules and approvals remain in control even when providers change.
Provider-neutral boundary
The current runtime exposes provider/framework-neutral ExecutionRequest, ExecutionPlan and ExecutionRun contracts. OpenAI-compatible execution is integrated; additional provider adapters remain capability expansions until they execute through the same governed path.
Context and writes
AI receives authorised context and can produce durable outcomes. Consequential writes can require human approval and are rechecked against authority before execution.
Truthful provider state
The Model Gateway tracks provider registry state so unavailable adapters are not presented as integrated capabilities.
Where this description comes from
These pointers are the authority behind the status and technical description. If implementation and documentation disagree, the canonical implementation or Control contract wins and this page should be corrected.