Enterprise Agent Fabric
An operating model for acting AI.
Identity, routing, policy, and proof run once. Domains keep the intelligence. This is not a chatbot framework. It is the shared front door and control plane for AI that does something — money, customer data, regulated process.
For business
What it governs
Who was allowed to use this. What it was permitted to do. Why this agent ran, and what it actually did. Those answers come from one place.
Start here
How it hangs together
The whole system in one sitting. Same spine for a CIO and a Chief Architect. Two depths after that.
For architects
How a request moves
Verify, route, pin, execute, record. Chat and jobs are the same idea. Only how the path is chosen differs.
What this is not
A portfolio of chatbots. A framework you drop into every domain. A place where Legal, payments, and KYC each invent identity, approval, and audit again.
People and systems never dial those URLs. They come in one door. The fabric verifies, routes, runs, and records. The domain AI does the work.
Four questions, answered once
Every acting capability needs the same answers. The fabric owns them so a domain team does not rebuild them.
| Question | What answers it |
|---|---|
| Who was allowed to use this? | Identity and entitlement at the door |
| Why this agent? | Routing against a versioned catalogue |
| What was it permitted to do? | Autonomy mode plus pinned tools and, when required, a human gate |
| What did it actually do? | Audit on one correlation_id |
Where to go next
| If you are | Open |
|---|---|
| New to the fabric | How it hangs together |
| Deciding who owns what | Operating model |
| Tracing a request | Request lifecycle |
| Choosing how much latitude the model gets | Autonomy |
| Looking at documents, review, or handoff | Use cases |
| Running the local stack | Start the fabric |