Manage parallel AI coding agents from one dashboard

Run multiple AI agents in isolated worktrees, monitor their output in real-time, manage PRs, and keep a separate mobile-friendly chat UI handy for checking in on worktrees away from your desktop.

$ bun install -g webmux
Used by engineers at Windmill every day.

Everything you need

A complete toolkit for running parallel AI coding agents.

Parallel AI Agents

Run multiple agents in isolated worktrees simultaneously. Each gets its own branch, terminal, and environment.

One-Click Worktrees

Create worktrees with AI-powered branch naming. Describe the task, and webmux handles the rest.

Real-Time Terminals

Live WebSocket terminal streaming in the browser. Watch agents work, send prompts, and switch instantly from the full desktop dashboard.

Mobile-Friendly Agents Chat

A separate agents UI gives you a much better mobile UX: worktree list, simple chat screen, and quick PR or service badges without trying to squeeze terminals onto a phone.

PR & CI Integration

PR status, CI check results, and review comments displayed inline next to each worktree.

Linear Issues

Browse and assign Linear issues directly. Auto-link branches to issues for seamless tracking.

Docker Sandboxing

Isolated containers with automatic port forwarding and volume mounts for untrusted workloads.

Create & Manage Worktrees

Spin up new worktrees with one click. Pick a profile, type a prompt, and webmux creates the worktree, starts the agent, and begins streaming output. Merge or remove worktrees when you're done.

PR, CI & Comments

See pull request status, CI check results, and review comments right next to each worktree. No more switching to GitHub to check if your agent's PR passed CI.

Service Health Monitoring

Track dev server ports across worktrees. webmux polls configured services and shows live health badges so you know which worktrees have their servers running.

Linear Integration

See your assigned Linear issues alongside your worktrees. Webmux matches branches to issues automatically, so you can browse your backlog, pick an issue, and spin up a worktree for it in one click.

One config file. Full control.

Define worktree layout, service ports, agent profiles, and Docker sandboxes in a single YAML file.

.webmux.yaml
name: My Project

workspace:
  mainBranch: main
  worktreeRoot: ../worktrees
  defaultAgent: claude
  autoPull:
    enabled: true
    intervalSeconds: 300

services:
  - name: API
    portEnv: PORT
    portStart: 3000
  - name: Frontend
    portEnv: FRONTEND_PORT
    portStart: 5173

profiles:
  default:
    runtime: host
    panes:
      - id: agent
        kind: agent
        focus: true
      - id: web
        kind: command
        split: right
        cwd: repo
        workingDir: frontend
        command: FRONTEND_PORT=$FRONTEND_PORT bun run dev

  sandbox:
    runtime: docker
    image: ghcr.io/your-org/your-image:latest
    yolo: true

Get started in 60 seconds

Four steps to go from zero to a running dashboard.

1

Install prerequisites

Install Bun and the required local tools webmux checks during setup.

sudo apt install tmux python3
curl -fsSL https://bun.sh/install | bash

On macOS, install tmux and python3 with Homebrew instead.

2

Install the CLI

Install the global webmux binary with Bun.

bun install -g webmux

This places the webmux CLI in your global Bun bin directory.

3

Initialize your repo

Create a starter .webmux.yaml in the repository you want to manage.

cd /path/to/your/project
webmux init

webmux init requires a git repository and verifies git, bun, python3, and tmux before setup.

4

Start the dashboard

Launch the local dashboard server for the current project.

webmux serve

The terminal dashboard runs on http://localhost:5111 by default, and the separate agents chat UI runs on http://localhost:5112. That second UI is primarily intended as a better mobile experience. Add --app for Chromium app mode or use --port / --agents-port to override the ports.