Skip to content
Agent-Fabric Agent-Fabric Agent-Fabric

Quickstart

Quickstart

Section titled “Quickstart”

TL;DR: install → sign in → connect → publish.

Terminal window
curl -LsSf https://raw.githubusercontent.com/falcons-eyes/agent-fabric-docs/main/install.sh | sh
fabric up
fabric serve http://127.0.0.1:11434 --name my-model --kind llm

That’s it — a machine joined to your private mesh, with a local model server published as an addressable, capability-gated service. The rest of this page walks through each step and what it actually does.

fabric — Onboarding
The flow this page walks through, as a real recording — sign in, join the mesh, discover a local model, publish it.

1. Install

Section titled “1. Install”

macOS / Linux:

Terminal window
curl -LsSf https://raw.githubusercontent.com/falcons-eyes/agent-fabric-docs/main/install.sh | sh

Windows (PowerShell):

Terminal window
irm https://raw.githubusercontent.com/falcons-eyes/agent-fabric-docs/main/install.ps1 | iex

This installs three binaries: fabric (the CLI you’ll use), afd (the background tunnel agent), and aflocal (a local dashboard — see Local Agent Stack). Verify it worked:

Terminal window
fabric version

2. Sign in and connect

Section titled “2. Sign in and connect”
Terminal window
fabric up

One command does three things: signs you in (opens your browser, or prints a device code if there’s no local browser), joins this machine to your network (generates a WireGuard key that never leaves the machine, and fetches a signed network map), and brings the encrypted tunnel up. Check it worked:

Terminal window
fabric status

3. Publish a service

Section titled “3. Publish a service”

Make anything running locally — a model server, an MCP tool, an internal API — reachable by your team over the private mesh:

Terminal window
fabric serve http://127.0.0.1:11434 --name my-model --kind llm

It becomes my-model.<this-machine>.private, visible under Services in the cloud console, and reachable by any teammate’s machine through a scoped, expiring capability — never a public port. Grant and resolve it from another machine:

Terminal window
fabric grant llm://my-model
fabric resolve my-model --cap <token>

Where to go next

Section titled “Where to go next”
  • Connect a device — the console side of onboarding: seeing your machines, networks, and services.
  • Local Agent Stack — running and verifying local model servers before you publish them.
  • CLI reference — every fabric command and flag.
  • How it works — the mesh + control-plane model, for anyone evaluating whether to trust it with their infrastructure.