v0.5.0Open source desktop app, MIT

Run coding agents in parallel, wherever they run best.

enowX is a lightweight, open source orchestrator for Claude Code, Codex and 43 more agents. Each works in its own git worktree, on your machine or your own server, and you watch, compare and merge from one native app.

  • macOS, Linux and Windows
  • MIT licensed
  • Local-first, no telemetry
esc
↑↓ select↵ open⌘K focus45 agents · macOS, Linux, Windows

The app has the same palette (Ctrl+K) for every session, host and action. Pick a command to go to that part of the page.

New session

Run agents in parallel

Every agent on its own branch, at once.

Each session is an agent in a real terminal, on a branch of its own in a git worktree of its own, so agents never overwrite each other or your checkout.

In a terminal, or a chat

Claude Code and Codex also show as a chat: their own transcript, with tool calls and diffs folded in, and a message box with the model, permission mode and effort. The terminal keeps running behind it.

45 agents built in

Each started the way it takes a first prompt, with its permission flags and its folder trusted. Any other terminal program runs too.

Claude CodeCodexGemini CLIOpenCodeKiro CLI+40

Status from the agents' own hooks

Notifications, unread marks and a sound of your choice. Idle agents hibernate to free memory, or you hibernate one yourself, and come back on their conversation.

workingneeds youdone

Side by side

CtrlShiftL

Several sessions in one view, each with its terminal, chat and browser.

Agent accounts

Several accounts of Claude Code or Codex on a host, a work and a personal one say, picked per session with no new sign-in.

Enter GOD of Agents

GOD of Agents

One session that drives all the others.

Say what you want to one session. It splits the work into tasks and hands each to a session on any host, with the agent and model the task needs, then follows them, checks what came back and reports. It reads files and writes what is small itself; the coding goes to its workers.

Delegates, follows, reports

Recorded in the app, with sample projects.

A model per task

The judge picks each prompt's model and effort, or a cheaper provider's config, and says why. Near a budget, cheaper ones first. Change model overrides it.

Your call on what matters

The judge and GOA answer workers' permission prompts. What can't be undone, spends money, publishes or touches secrets comes to you with a sound and a banner.

Rules in your own words

Say how it should work, in any language. An AI writes it as clear rules for you to read and save, followed on every host.

Any agent, held as far as it allows

On Claude Code and Codex, hooks refuse what it shouldn't run. Other agents are watched, and what they change is flagged. Each agent's level shows when you pick one.

enforcedwatched

A board and a memory

Each task, its session, status and result, beside its chat. Outcomes, decisions and your preferences stay in a RAG collection of its own.

Search by meaning

RAG

Agents find the code without reading it all.

enowX indexes a project itself, and agents search it by meaning and by words: its code, docs, attached notes and the decisions kept as memories. A project is indexed only after you agree to send its text to the embedder you chose.

Index, search, open

Recorded in the app, with a sample project.

Consent first

A dialog says which service gets the text, about how many tokens, and what stays out. Secrets and ignored files never go.

Your store, your embedder

One SQLite file on the host, or PostgreSQL with pgvector, your own server if you like.

VoyageOpenAI-compatibleTEI

Kept up to date

As sessions change files, in the background, unchanged parts skipped. Parts that bear on a prompt can join it as context, and a session's decisions can be kept as memories.

For every agent

Also as MCP tools
enowx-app rag search "where is tax added"enowx-app rag remember "VAT is charged at checkout"
Open Decisions

Decision models

A small model answers, so nothing has to guess.

Is a quiet session waiting for you? Is this command safe to run? What of this long output does the agent need? enowX asks a small, cheap model, the judge, and acts only when it is sure enough. A failure never blocks an agent.

Decisions in the app

Recorded in the app, with a sample local model.

What it decides

  • Allow a command, leave it to you, or refuse it
  • Trim long output to what matters
  • Stop a loop before more tokens go
  • Pick a prompt's model
  • A flaky or a real test failure

What can judge

Found on the host: agents you signed in to, model APIs, local servers, and decision models made for this.

claude -pcodex execOllamaJevClefNeuralwatt

Your policy

Only the session's own provider, one provider, any, or local only so nothing leaves the host. Each use has its switch, how sure it must be and a daily limit.

A log to correct

Every decision with its verdict, reason, model and time. Mark what was wrong, and see what enowX did about it.

Open Hooks

Hooks, token usage, configs

One set of hooks on every agent, and every token counted.

Hooks that run the same on any agent, what each agent spends, and other accounts and model APIs per session, without touching an agent's own config.

Hooks on every agent

Installed into an agent's own hooks, or detected from outside for agents without them. Events for sessions, prompts, tools, files, Git and schedules; conditions on agent, project, branch, tool and text. A project's hooks don't run until you allow them.

runadd contextrefuserewritejudgenotifywebhook

Token usage

Each agent, model, project, session, config and host, per day: input, cached, output, costs and budgets. Only counts, never prompts or answers, kept by each host, with no telemetry.

Configs per session

Claude Code on a third-party model beside the official one, picked at New session and shown as a badge. An agent's own files are never written.

Fallback at a limit

At a limit or out of credits, a session goes on with the next account or provider in your order, automatically or after asking.

Fan out

Fan-out, pipelines, automations

One prompt, several agents. Keep the best.

Send one prompt to several agents, or the same agent twice, then compare what each changed and keep the best. The others are discarded with their worktrees and branches.

Fan-out

The same task, in parallel, each agent in its own worktree.

Compare

What each one changed, file by file, and a button to keep it.

Compare: two sessions given the same prompt, each with the files and lines it changed, and a button to keep it.

Pipelines

Plan, implement and review a task, agents one after another, each stage starting from what the last ended with.

  1. 1Plan
  2. 2Implement
  3. 3Review

Automations

A prompt on a schedule, run by the host's daemon whether the app is open or not.

  • Weekdays: triage new issues
  • Nightly: update the dependencies
Review changes

Review and ship

Read every change before it lands.

The Git tab shows what a session changed against its base, file by file. Send comments back to the agent, commit, push, open a pull request and merge, or discard the worktree.

The Git panel

Commit and merge

Commit, push, a pull request with gh, then merge.

Conflict radar

Warns when another session changed the same files, and on which lines. Sessions can start from a GitHub issue.

The side panel

The tools beside every session.

GitChecksFilesPortsActivityBrowserTrafficAndroid
Open the Browser

See it running

The dev server, the desktop app, the phone.

Whatever the session builds, you see it running, wherever the session runs: on this machine or on your server.

Browser

The session's dev server in a native web view, with hot reload, an address bar, a design mode to point at an element and hand it to the agent, and screenshots. Agents can drive it themselves.

Resource usage

Memory and CPU of every session, counting everything it started: language servers, MCP servers, dev servers.

Resource usage: memory, CPU and disk of the host, and of each session.

Desktop apps from a server

An agent builds a Tauri, Electron, Flutter, GTK or Qt app on your server, and it opens on your laptop in a window of its own, with the framework's own hot reload.

  • A virtual display per session, with Xpra
  • Keyboard, pointer and clipboard go back to the server
  • Linux servers

Android

Runs the session's app on an emulator or a phone, on its host or on your laptop, shows its screen with scrcpy and its log, and catches its traffic.

FlutterExpoReact NativeCapacitorNativePHPKotlinJava
Open Canvas

Canvas

Design with agents, on an endless board.

Mockups in HTML, wireframes, diagrams and boards on each host. Watch an agent build a design element by element, or ten variations to compare and star, then have it turn the one you pick into code.

The Canvas page

Scroll to move the board and pinch to zoom, as in a design tool. Move and edit shapes, pause or stop the agent, restore any version.

The Canvas page: a checkout flow drawn as a diagram, its other designs in a strip beside it.

Many designs

  • Numbered, grouped by page and variant
  • All of them in a grid, or two to four compared
  • Starred and picked, which agents read

Click a component

Ask the agent about it, have it changed or redesigned on its own, or leave a note on it.

The agent beside it

Claude Code, Codex or another agent, held to the canvas's tools. In review mode, keep or undo what each message changed.

The canvas with Claude Code's panel beside it, offering variations, responsive frames, a critique or a restyle.

Out of the canvas

Export a design, share a read-only link on this device and your tailnet, or have an agent build it in a project, in its stack and components.

PNGSVGPDFHTMLLayout gridsMake code
Open Notes

Notes

Notes your agents read, and help write.

Pages of notes on each host, written a block at a time with Markdown's shortcuts and / for blocks. Every version is kept to restore.

Blocks, not only text

Tabs, toggles, columns, callouts and cards, to-dos with their progress, an outline to jump through, and SVG icons in any color.

The Notes page: a launch plan with a callout, to-dos and their progress, and two columns.

Ask an agent

Beside a page, Claude Code, Codex or another agent tidies it, plans it or brainstorms it. Its changes are yours to accept.

Attached to projects

A note attaches to projects on any host, and an agent working there reads it.

Import and export

Bring notes in with their images, from Notion again on a schedule. Export to Markdown, folders, HTML, PDF or a backup.

NotionObsidianEvernoteKeepHTML
Open Credentials

Credentials and your data

Your logins on every server, never in an agent's hands.

A credential vault encrypted on this device, lent to a session on any host only when you allow it. git, gh, docker and ssh on a server then work without signing in there.

The credential vault

GitHub, clouds, hosting, databases, registries and AI APIs. Agents use a credential by name, and its value is masked in what they read.

The Credentials page: GitHub, Hetzner, Cloudflare, Neon and npm logins, each with the end of its value and a Test button.

Filled without the command line

  • From what the host's tools already keep
  • A token copied on the platform's page
  • Signing in in the browser

Your data on every host

Notes, canvases, settings and the vault, encrypted, are the same on this device and every server, merged when changed on two at once.

Features on demand

Turn what agents are told about on or off, everywhere or per host, project or session, so their prompts stay short.

Settings, Features: what the agents' guide costs in tokens, presets, and a switch for each feature.

Sealed on this device

AES-256 under a key from your master password, kept in the system's keychain if you like. Hosts keep only the ciphertext, and the vault locks after 15 minutes unused.

Open Traffic

Traffic

See and shape an app's HTTP.

Traffic shows what the apps a session runs send and receive, on its host, live, over TLS too: requests, responses, WebSocket, server-sent events and gRPC.

Attach what you want caught

Each host has its own certificate authority. Nothing is caught until attached.

A terminalThe BrowserChromeFirefoxElectronDockerAndroid

Mock, forward, change, slow, fail or pause

Rules kept in the project, in .enowx/traffic.json: a mock from a request in one click, breakpoints to edit a request or a response before it goes on.

Inspect and compose

Headers and bodies by kind, timings and TLS, filtered by domain, status, type, size and time. Pause or stop recording when the list gets busy. Copy a request as curl, fetch, Python, Go, HTTPie or PowerShell; export and import HAR.

Agents use it too

From the command line, and the same as MCP tools, secrets masked.

shell
enowx-app traffic listenowx-app traffic watchenowx-app traffic mockenowx-app mcp
Add a host

Your laptop is just the screen

Agents run on your server. The laptop shows them.

The app installs its daemon on your server, where it runs as a systemd service: sessions keep running with the laptop closed and come back after a reboot. Over SSH, or Tailscale.

One-click setup

A blank server gets what this device has, in one step.

  • System packages
  • Toolchains at the same versions
  • git and gh
  • The agents and their sign-in

Create a server

At Hetzner, DigitalOcean, Vultr or InstaBox, from the app, with your own account and token, or an InstaBox machine you have.

HetznerDigitalOceanVultrInstaBox

Sync environment

Skills, subagents, commands, memory files, settings and MCP servers, between this device and a server, either way.

Copy a project, sync it back

A project and its sessions to another host, then files, commits and conversations back. Copies can also sync on their own, and push the work to a branch of its own.

Far servers stay connected

Connections come back by themselves and keepalives keep them open. Tailscale is built in.

Team

Work as a team

Your team, around the same projects.

enowX Cloud is opt-in: sign in with GitHub, make a team or join one, and share notes, projects and hosts with it, end-to-end encrypted. Nothing reaches the cloud until you sign in.

Team chat

In the side panel, beside your sessions: @ mentions members, notes, canvases, projects and files, and messages edit and delete. The panel also lists the team's projects, where each is and the sessions in it, and who is on your host.

Invite by link

The link works in the browser too: someone without the app signs in with GitHub and joins there, then downloads enowX and finds the team.

OwnerAdminMemberViewer

Shared hosts

Share a host with the team over Tailscale. Members watch a session live, ask to take it over, or start their own with their own agent account, never yours.

A team's space

What the team keeps together, beside each member's own.

  • Notes and canvases
  • Projects as Git links, cloned in one click
  • A kit of skills, pipelines and templates
  • Each project's rules for the agents

End-to-end encrypted

What a team shares is sealed with the team's key on its members' devices: the cloud holds ciphertext only, and credentials never go to it. An activity log and a recovery key come with it.

enowx-app

Script it

Agents that start other agents.

The enowx-app command line talks to the host's daemon, for your scripts and for agents that start other agents.

The command line

shell
id=$(enowx-app spawn --agent codex --branch fix/parser "Fix the parser's handling of empty lines")enowx-app wait "$id"      # until the agent is done or asks somethingenowx-app read "$id"      # what its screen showsenowx-app send "$id" "Now add a test for it"enowx-app spawn --agents claude,codex "Add a search box"   # a fan-outenowx-app pipeline "Paginate the notes list"               # plan, implement, review

And more

Also, and browser to drive a session's Browser:

lsstatusnotehibernateprojectspipelinesautomationbrowsercanvasnotescredssyncfeatures
Change themeCtrlK

Make it yours

Five themes, any accent, one palette.

A command palette for every session, host and action, five themes and any accent color, and a floating control over every app.

Themes and the palette

CtrlK

Floating control

CtrlShiftSpace

A round panel over other apps' windows: a new session, a folder, the palette, and the sessions of this PC or a server.

The floating control: Sessions, New session, Folder, Panels, Search, Settings and Appearance around the enowX button.

Sounds

A sound for an agent done, waiting for you or failed, built in or your own, with quiet hours.

Settings, Notifications: a sound for each event, the volume, quiet hours and a count on the app's icon.

Settings

  • The agents' library: skills, MCP servers, subagents, commands, instructions, plugins
  • What new sessions start with
  • Environment variables every session gets
  • Automations and diagnostics

Updates itself

With signed archives, or a delta from the running version.

Local-first

No account needed, no relay, no telemetry. The app talks to your machine and your servers directly; the team cloud is used once you sign in, for what you share. Passwords and tokens stay in the keychain and the encrypted vault.

Browse agents

Agents

45 agents, and any CLI you add.

Claude Code and Codex are the furthest integrated. Several accounts of the same agent work side by side.

  • Claude Code
  • Codex
  • OpenCode
  • Gemini CLI
  • Kiro CLI
  • Claude Agent Teams
  • OpenClaude
  • Grok
  • GitHub Copilot
  • OpenCode 2
  • MiMo Code
  • Ante
  • Trae
  • Muse
  • DeepSeek Harness
  • Qoder CLI
  • Qoder CLI China
  • ZCode
  • Pi
  • OMP
  • Prime Agent
  • Antigravity
  • Aider
  • Goose
  • Amp
  • Kilo Code
  • Crush
  • Auggie
  • Autohand Code
  • Cline
  • Codebuff
  • Freebuff
  • Command Code
  • Continue
  • Cursor
  • Droid
  • Kimi
  • Mistral Vibe
  • Qwen Code
  • Rovo Dev
  • Hermes
  • Devin
  • OpenClaw
  • CodeBuddy
  • Jcode
  • Any terminal program

Download

Run your agents in parallel.

One native app for macOS, Linux and Windows. Open source under the MIT License, with no account and no cloud.

Version 0.5.0 · released 1 Nov 2026 · Release notes

Files come straight from GitHub Releases.