Foxwarm

Self-hosted agent workspace

Keep the thread.
Own the workspace.

Foxwarm keeps each Agent's Sessions, tools, and memory in an installation you control.

Install FoxwarmRead the docs
curl -fsSL https://foxwarm.550w.host/install-foxwarm.sh | bash

Requires Git, Node.js 20+, npm, and tmux. Windows has a native PowerShell installer.

How Foxwarm is organized

Agents, Sessions, tools, and models

Foxwarm connects Agent memory, Session history, tools, and model access while keeping their roles clear.

01

Sessions that carry on

Sessions survive restarts. Foxwarm compacts long conversations into traceable layers and can recall archived source when earlier context matters.

02

One Agent, several threads

An Agent holds its workspace and long-lived memory. Several Sessions can share that Agent while keeping their conversations separate.

03

Tools close to the work

Use built-in file and shell tools, load reusable Skills, connect MCP servers, or run tools on another environment through a Node.

04

Use your model provider

Connect a supported OpenAI-compatible or Anthropic provider. WebUI works on its own, and messaging Channels are available when your workflow needs them.

Your first Session

Install, connect a model, and start a thread

  1. 1
    Install

    Run the installer on Linux, macOS, WSL, or Windows. Program files and runtime data stay separate.

  2. 2
    Choose a model

    Open the tokenized local WebUI and finish first-run setup with your provider, model ID, and API key if required.

  3. 3
    Start working

    Use the default Agent and Session, load a Skill when a task needs it, or create a focused thread. Add Nodes and Channels later.

Install locally

Run Foxwarm on your machine.

Begin with WebUI and one model. You can add Nodes and Channels later.

Installation guideView source on GitHub