Playbooks
Versioned rules that define an expert's behavior.
Product · Primitives
The building blocks for governed AI work.
Use these primitives together to build an agentic expert, govern every turn, retain what matters, and make the work visible.
01 · System layer
Engine
Governed control and execution
Protocols
Ordered stages that make up a governed turn.
Routing
Selects the permitted model, provider, and capability.
Observer
Extracts the structured markers a policy needs.
Tools
Authorizes the external capabilities a turn can use.
AISEC
Records receipts, decisions, reasons, codes, and evidence.
Credits and ledger
Reserves, settles, releases, and records every approved cost.
02 · Workspace layer
Workspace
Owned knowledge, work, and collaboration
PRIME
Brings governed capabilities into a user's workspace.
Workspace
Defines ownership, membership, and the permission boundary.
Projects
Connect a goal to its knowledge, conversations, and work.
Knowledge Base
Retains visible organizational knowledge outside chat history.
Context and state
Keeps provider-neutral memory and current work context.
Artifacts
Creates owned outputs with lifecycle and provenance.
Work
Turns conversations into goals, todos, and evidence.
Learning
Retains learning derived from completed work and artifacts.
People and groups
Gives teams a shared collaboration and ownership surface.
Manifest access
Controls which expert playbooks a workspace may use.
03 · Capability layer
Provider catalog
Approved external model capabilities
Provider catalog
Lists approved model and provider capabilities with prices.
Build with the primitives