What is Wexa Fabric
Wexa Fabric is the governed context layer between your enterprise data and the agents acting on it. It sits between your data estate and your AI agents, so every agent is grounded in shared context and every action it takes is policy-checked, simulated, approved, traced, and audited before it touches production.
Instead of wiring each agent directly to your systems — and rebuilding connectors, guardrails, and audit for every team — you connect your agents to Fabric once. They inherit governance structurally, simply by routing through it.
What Fabric gives you
Section titled “What Fabric gives you”- One context layer — a shared, self-organizing context graph and graph-aware retrieval that ground agents in your real enterprise data and the relationships within it.
- A single endpoint, three ways to call it — connect over MCP, REST, or the Python / TypeScript SDKs. Every call is scope-pinned to your project, policy-checked, and audited.
- Governed action — connectors expose read and write operations as governed tools; high-risk actions are simulated and human-approved before they run.
- Open interoperability — Fabric is an MCP server to inbound agents and an MCP / A2A client to outbound agent platforms, so you are never locked to one ecosystem.
- Deploy anywhere — the identical platform runs as managed cloud, in your private cloud / VPC, or fully air-gapped — selected by configuration.
Who Fabric is for
Section titled “Who Fabric is for”Fabric serves the people who run AI in the enterprise — IT, data, security, and governance teams — through a single, role-aware administration console. There is no ungoverned end-user surface: every call is scope-pinned, policy-checked, and audited, whether it comes from the console, the API, or an agent.
Fabric offers two front doors:
- A streamlined connect-and-go experience (Simple Mode) — one page to point an agent at the platform: pick a workspace, get ready-to-use MCP credentials, and copy a REST or SDK quickstart.
- The full governed console (Advanced Mode) — the complete control surface for administrators: every capability domain, the organization / department / project scope trail, and every trust surface.
The streamlined mode removes approval friction for low-risk workflows; it never removes governance.
The two directions of MCP
Section titled “The two directions of MCP”It helps to be precise about the word “MCP”, because Fabric uses it in both directions:
| Direction | What it means | Where you start |
|---|---|---|
| Fabric as an MCP server | Your agent (Claude, Cursor, ChatGPT, …) connects into Fabric to query your context graph | Connect Fabric as an MCP server |
| Fabric as an MCP / A2A client | Fabric reaches out to external tools and agent platforms on your behalf | Via interop agents — see Create an agent |
This documentation focuses first on the inbound direction — installing Fabric as an MCP server inside your agent — because that is the most common way to start.
Next steps
Section titled “Next steps”- Core concepts — the vocabulary you need.
- Architecture at a glance — how the pieces fit.
- Quickstart — connect an agent in five minutes.
