Agent Fabric

A governed capability layer that routes work through skills, tools, and audit boundaries instead of letting one model improvise everything.

An agent is only useful when the system knows what it can do, what it must not do, and how to prove every action after the fact.

Capability routingTool scopingHuman review gates

What this proves

Separate skill selection from tool execution so context stays narrow.

Keep MCP connectors explicit, named, and auditable.

Make every action traceable back to a human-requested intent.

Open-source stack

MCPSkillsFastAPIPostgresOpenTelemetryPolicy checks

Experience mode

Step 1

Intent

Step 2

Skill loader

Step 3

Tool contract

Step 4

Result

Task framing

The system becomes easier to trust when capability is explicit instead of implicit.

Live pattern

Engineering lens

Skills define the minimum useful context for the job.
The system can load only the domain needed for the task.
The user sees a clear handoff from request to output.

Platform fit

This project belongs in Llewellyn Systems because it turns a repeated engineering pattern into a governed operating asset. The page is not a slide deck. It is a proof surface for how the system is built and how it behaves.

Toolchain note

Use JupyterBook for publication, MyST for source text, Voilà for notebook apps, Binder for reproducible environments, and JupyterLab or Colab for interactive editing. The page itself is the front door to that workflow.

Related projects

JupyterBook + MyST