WatchfireWatchfire

Better context. Better code.

Define what you want. Let Claude Code, Codex, opencode, Gemini CLI, and GitHub Copilot CLI build it — safely.

Watchfire turns clear specs into scoped tasks, then hands each one to an agent in its own git worktree and sandbox. Better context in, better code out — with clean transcripts and full control at every step.

How it works

Two ways to ship code

Full autonomy, or full control — pick whichever fits the job.

1

Define your project

Run watchfire init, pick your agent backend, and describe what you want to build — Watchfire generates the project definition and initial tasks

2

watchfire wildfire

Launch Wildfire mode — a fully autonomous loop that runs without human intervention

3

Execute

Auto

Agents pick up ready tasks and work on them in isolated git worktrees, one branch per task

4

Refine & generate

Auto

Once tasks are done, the agent refines draft tasks and generates new ones — then loops back to execute

5

Auto-merge

Auto

Completed work is automatically merged back — Wildfire keeps going until there's nothing left to do

Architecture

Three components, one system

A daemon orchestrates in the background while the CLI and GUI give you full control.

watchfire — tuiTASKSOUTPUTready? help

CLI / TUI

watchfire

Project-scoped commands plus a rich interactive TUI with split-pane task navigation and live output.

running

GUI (Watchfire.app)

Watchfire.app

Electron multi-project client with a native feel, sidebar navigation, and embedded live terminals.

ProjectsAgentsGitgRPC

Daemon (watchfired)

watchfired

Background service handling orchestration, PTY management, git workflows, and the gRPC API.

New in v3.0.0

Any agent. One orchestrator.

Watchfire is backend-agnostic. Plug in your favourite CLI coding agent — or mix and match across tasks.

C

Claude Code

Anthropic

$claude-code
Cx

Codex

OpenAI

$codex
oc

opencode

Open source

$opencode
G

Gemini CLI

Google

$gemini
Co

Copilot CLI

GitHub

$copilot

Project default

Pick one agent per project during watchfire init — or change it anytime from the settings tab.

Per-task override

Pin individual tasks to a specific backend — Claude Code for architecture, Codex for trivial edits, or swap agents without touching project settings.

Unified transcripts

Every backend renders into the same clean User/Assistant log — so you review a session the same way no matter which agent ran it.

Modes

From pair programming to autonomous loops

Six modes that cover interactive chat, single tasks, and fully hands-off Wildfire runs.

Flagship

Wildfire

Autonomous loop — execute ready tasks, refine drafts, generate new tasks, repeat.

Chat

Interactive session with the coding agent — ask questions, iterate, and pair-program.

Task

Execute a single task from the task list in its own worktree.

Start All

Run every ready task sequentially, one worktree per task.

Generate Definition

Auto-generate a project definition by scanning your codebase.

Generate Tasks

Break down the project definition into discrete, ready-to-run tasks.

Key features

Built for real agent workflows

Isolation, sandboxing, and orchestration — the plumbing that makes multi-agent safe.

01Orchestration

Task Orchestration

Ship work without babysitting sessions. Write tasks in YAML and the daemon handles the rest — picking up ready work, stopping sessions on completion, and chaining to the next task. Every run lands in the same clean transcript, no matter which backend produced it.

donerunningready
02Isolation

Git Worktree Isolation

Parallel agents, zero merge conflicts. Every task runs in its own git worktree on a dedicated branch, so agents work side-by-side without touching each other's files. Finished work auto-merges back to your default branch.

mainwatchfire/001mergedwatchfire/002runningwatchfire/003QUEUED
03Security

Sandboxed Execution

Your credentials and dotfiles stay off-limits, even if an agent goes off-script. Every session runs inside a platform-native sandbox — Seatbelt on macOS, Landlock or Bubblewrap on Linux — blocking ~/.ssh, credential stores, and git hooks by default. Restart protection stops runaway loops before they burn budget.

SEATBELT · LANDLOCK · BUBBLEWRAPSANDBOXagent/0042$ watchfire run→ loading spec→ worktree ready✓ sandbox active$ building...HOST SYSTEMworkspace/READ · WRITE.watchfire/READ · WRITE~/.ssh/BLOCKED~/.aws/credentialsBLOCKED.git/hooks/BLOCKED
04Scale

Multi-Project Management

Drive every repo on your machine from one place. A single daemon runs multiple projects in parallel — each with its own task queue, worktrees, and agent sessions — and the system tray keeps them all one click away.

watchfireddaemon
Ready to go

Get started with Watchfire

Install in seconds. Define tasks. Let agents ship code for you.

Download for macOS

Available for macOS, Linux, and Windows

install
# Install via Homebrew (macOS)
$brew tap watchfire-io/tap
$brew install --cask watchfire-io/tap/watchfire
# Set up your project and go
$watchfire init
$watchfire task add "Build the login page"
$watchfire start --all

Includes GUI, CLI, and daemon. Also available via Homebrew.