The Open Source Drop
ToolsExploreAlternativesThis Week
Subscribe
The Open Source Drop
ToolsExploreAlternativesThis Week
Subscribe
← All Categories

Coding Agent Workspaces

14 open source tools compared. Sorted by stars. Scroll down for our analysis.

By Erik Loyd, SaaS CEO and former COO/CFO of an AWS Premier Partner.

ToolStarsVelocityLanguageLicenseScore
orca

Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop and mobile.

126.1k+7325/wkTypeScriptMIT License93
multica

The open-source managed agents platform. Turn coding agents into real teammates: assign tasks, track progress, compound skills.

56.3k+1020/wkTypeScriptMultica License (Apache 2.0 + conditions)79
cmux

Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents

28.8k+50/wkSwiftGPL-3.0-or-later (app) + BUSL-1.1 (server)85
Superset (superset.sh)

Code Editor for the AI Agents Era - Run an army of Claude Code, Codex, etc. on your machine

16.3k+214/wkTypeScriptElastic License 2.076
agent-orchestrator

Run and supervise teams of coding agents from planning to merge. Any harness (Claude code, codex, +25 more). Desktop, web, mobile, and cloud agents.

14.8k+469/wkGoApache License 2.090
omnigent

A meta-harness for all your AI agents. Omnigent provides a common layer over Claude Code, Codex, Pi, and the agents you write yourself: swap or combine harnesses without rewriting, keep them in check with policies and sandboxing, and collaborate in real time on the same live session, from any device.

11.4k+94/wkPythonApache License 2.084
cindy

Consider it done. The open-source AI agent that works out of the box · 想到,就能做到。开源、开箱即用的 AI Agent。

10.2k+723/wkTypeScriptApache License 2.084
superplane

Open source factory for one-shot engineering

10.1k-GoApache License 2.068
coucou

A tiny friend that lives in your notch (macOS) or at the top of your screen (Windows) and keeps an eye on your Claude Code sessions.

7.3k+188/wkSwiftMIT License78
agent-autoloop

让你的命令行 Agent 在后台自己接活、自己跟进到底;平时不烦你,遇到拿不准的事才发卡片找你拍板。

5.6k+1391/wkTypeScriptMIT License84
hapi

App for Claude Code / Codex / Gemini / OpenCode, vibe coding anytime, anywhere

5.4k+48/wkTypeScriptGNU Affero General Public License v3.072
zeron

A native control plane for Claude Code, Codex, Cursor, Devin and other coding agents.

3.7k+207/wkRustMIT License74
openharness

The ultimate harness for coding agents and beyond. All your agents. All your machines. One command center. Start with code, then follow your curiosity and build across disciplines: CAD, circuits, robots, games and music.

1.5k+49/wkDartMIT License70
herdrm

Native macOS console for herdr: all your coding agents and their live terminals, across devices

796+2/wkSwift-55

Stay ahead of the category

New tools and momentum shifts, every Wednesday.

Our Analysis

orca126.1k★

Orca runs several coding agents at once without them stepping on each other. Claude Code, Codex, opencode, or any CLI agent each gets its own git worktree, and you watch them in chat or raw terminal panes, review the diffs, and send line-level feedback back. Desktop app for macOS, Windows, and Linux. MIT licensed and free. Running it is a download. It never sits between you and the model: agents use the subscriptions and keys you already have. For always-on work, orca serve runs a headless server inside your network, and cloud VMs work if you bring the provider. A phone app on iOS and Android pings you when an agent needs input. No paid tier exists. Solo: free, and a workflow upgrade the first day. Small teams: free per person. Larger teams: Orca Enterprise is contact-only, and its security page claims SOC 2 readiness, not a completed audit. The catch: parallel agents multiply your token bill. Five agents on one task means paying for five attempts. Worth it on hard problems, wasteful when one agent would have nailed it. Usage telemetry comes with an opt-out, and the docs call it pseudonymous, not anonymous.

multica56.3k★

Multica turns AI coding agents into names on an issue board. Assign an issue to Claude Code, Codex, Cursor or any of 26 supported agent CLIs, and the agent picks it up, comments as it works, and hands the result back for review. Squads, reusable skills and scheduled autopilots sit on top. Self-hosting it for your own organization is free. Setup is one install script plus multica setup self-host, which pulls Docker images for a Go backend, a Next.js app and Postgres 17. Everyone who runs agents also installs a daemon on their own machine, where the work happens. Multica ships no model; you bring the agent CLIs and their subscriptions. The hosted cloud has a Free plan and a Pro plan billed per human seat (agents never take one), but the price isn't published; you see it at Stripe checkout. Solo and small teams: self-host free, or look at Vibe Kanban for a lighter local board. Larger teams: still free internally. The catch is the license. It's Apache 2.0 plus conditions: hosting it for anyone outside your company needs a commercial license even if it's free, the logo stays, and contributors agree the company can tighten the terms later.

cmux28.8k★

cmux is a native macOS terminal built on Ghostty's engine for running a pile of coding agents at once. Vertical tabs, split panes and a sidebar show each session's git branch, PR status, working directory, open ports and latest notification, and panes light up when Claude Code, Codex or Gemini needs you. The terminal is GPL-3.0 and free. Download the DMG or brew install --cask cmux; it's signed, notarized and auto-updates. It adds a scriptable browser pane that imports cookies and sessions from Chrome, Firefox, Arc and 20+ other browsers, SSH workspaces that route browser traffic through the remote machine, session restore, and one-command Claude Code teammate mode. An iPhone companion is in TestFlight. The terminal and local agents stay free. The money is in cloud agents: Pro is $50 a month for up to 5 cloud VMs sharing 20 vCPUs, Max is $200 for four times the compute, Team is $60 per user a month, and Enterprise adds self-hosted cloud execution, SSO and SOC 2. Solo devs on a Mac: free. Teams wanting agents running in the cloud: price Team against your own VMs. The catch: the desktop app is macOS only, and the license got more complicated in September 2026 when the server code moved to the Business Source License. The app stays GPL; the cloud side is source-available. It's young and changes fast.

Superset (superset.sh)16.3k★

Superset is a desktop app for running a pile of coding agents at once. Claude Code, Codex, Cursor Agent or any CLI agent each get their own Git worktree, terminals, diff viewer and browser preview. Not Apache Superset, the BI dashboard tool. The app is free and you bring your own agent subscriptions. Setup is a download. macOS is the main target, Linux gets an experimental AppImage, Windows nothing yet. Know what you're installing: it's an Electron app that adds hooks and command wrappers to your agent configs, and a worktree is not a sandbox. Agents still share your machine, network and files. Free is one user and local workspaces. Pro is $20 per user per month, or $15 billed yearly, and adds teammates, remote access, cloud workspaces, scheduled automations, the iPhone app, and Slack and Linear. Enterprise is quote-only for SAML SSO, SCIM and audit logs. Solo: free. Small teams: Pro. Larger teams: Enterprise. One agent at a time? Just use opencode or Aider. The catch is the license. Superset was Apache 2.0 until February 2026, then switched to Elastic License 2.0: source-available, not open source, no reselling it as a hosted service. Fine for internal use.

agent-orchestrator14.8k★

Orchestrator.inc runs a whole team of coding agents from one desktop app. Each task gets its own worker: one agent, one branch, one git worktree, so parallel agents stop trampling each other. A planning agent splits big goals into tasks, and a live Kanban tracks every worker through PR, CI and review. It drives Claude Code, Codex, Cursor, Aider, Goose and plenty more. Free, Apache 2.0. Setup is a download for macOS, Windows or Linux; the app runs its own daemon. You bring your agent CLIs and API keys, and it doesn't proxy model calls, so your model bill is the real cost. Best trick: failed CI and review comments go back to the agent that owns the branch. Nothing is paywalled for individuals. A hosted Cloud is waitlist-only with no published price, and teams can buy a design-partner pilot at $500-2,000/month for early shared dashboards. Solo and small teams: free. Larger orgs wanting team features: the pilot, for now. The catch: release builds send telemetry by default, including your GitHub username, and opting out takes three environment variables. It ships nightly with a deep issue backlog, so expect churn. Compare Conductor and Claude Squad.

omnigent11.4k★

Omnigent is one control layer over your coding agents. Claude Code, Codex, Cursor, OpenCode, Hermes, Pi and agents you define in YAML run through the same server, so you can mix them in one session, have one review another's work, or swap via config. Built by Databricks' AI team with Neon, Apache 2.0, free. Install is one curl script or uv tool install omnigent on Python 3.12+. Run the server locally, or docker compose up on a VPS and sessions follow you to browser, phone or desktop. Policies can pause for approval before shell commands or cap spend, and agents can run in Modal, E2B or Kubernetes sandboxes. No paid tier. You pay for the agents underneath, and three agents in one session means three token bills. Solo, small and medium teams: free. Large teams: wait for a stable release. The catch: this is alpha software whose job is containing agents, and its security record says that job isn't done. It has published 13 GitHub security advisories since June, four critical, including policy bypasses, host shell escapes and a default auth mode that allowed impersonation. All are patched as of 0.18.0, so stay current. Telemetry is on by default (OMNIGENT_ANALYTICS=0).

cindy10.2k★

Cindy is a desktop app that sits on top of the coding agents you already use. It drives Claude Code, Codex and Pi, switches harness or model mid-task while memory and skills carry over, and can split one task across agents to plan, build and review. XD Inc. makes it; the client is Apache 2.0 and free. Bring your own API key, a local model via Ollama, or the Claude Code or Codex plan you already pay for, with no second bill. "Skip Sign-In" runs local agents with no account, but server-backed features drop out, and the backend behind the phone remote and Slack delegation isn't open source. Paying buys model credit, not features: Plus is $20 for a month ($16 on auto-renew) with $20 of usage across GPT, Claude, Gemini and others; Super and Ultra are $100 and $200. Solo and small teams: free with your own keys, since Team controls are still waitlist-only. Larger teams: Enterprise (VPC gateway) is contact-sales. The catch: it's early, with well over a thousand open issues. Official builds ship TapDB usage analytics tied to your account ID once you sign in (no chat or file content). Build from source to drop it.

superplane10.1k★

SuperPlane is a software factory. It takes small, clear backlog issues, hands them to coding agents, checks the result against your CI and an agent review, and sends failures back until there is a pull request worth a human's time. Work comes in from GitHub, Jira, Linear or Sentry. The engine is Apache 2.0, and the community edition is free to self-host. Self-hosting the community edition means Docker Compose on one Linux VM plus runner machines. The Kubernetes install on AWS, Azure or Google Cloud needs an Enterprise license, even though the README lists it as a self-host option. Role controls and GitHub Enterprise support sit in a commercially licensed ee folder. The cloud starts at $0 plus usage, then $20 per organization a month (Starter) and $50 per user a month (Team). Solo: free, pay-as-you-go or one VM. Small teams: Starter, ten seats for $20 is fair. Larger teams: Team or Enterprise, and SSO and audit logs are Enterprise only. The catch: it only works on issues clear enough for an agent to one-shot, and most backlogs aren't. Sandbox time bills even with your own API keys. Two July CVEs were fixed before 1.0, so stay current.

coucou7.3k★

Coucou puts a little character in your MacBook's notch (or at the top of the screen on Windows and Linux) that watches your AI coding agents: which files Claude Code, Cursor, Codex or Gemini CLI are reading and editing, the diffs, and when they finish. You can answer permission prompts from the notch, from your iPhone's Lock Screen, or (paired with the Windows or Linux app) from an Android phone. The code is MIT and there's no paid tier. On a Mac, download the release, drag it to Applications, and click Install hooks; it backs up your Claude Code settings and shows the diff before writing anything. The Mac build is signed and notarized. The Windows installer isn't code-signed yet, so expect an "unknown publisher" SmartScreen warning the first time. Some agent integrations (Gemini CLI, OpenCode, Amp) are Mac-only. The iPhone app is a TestFlight beta, and the Android app has no signed release yet, so you build it from source. Solo devs running long agent sessions who want to step away from the keyboard get the most out of it. There's nothing for teams to buy or manage. If you only want a general notch utility, boring.notch (TheBoredTeam/boring.notch) is the established open source pick. The catch: approving agent actions from your phone is a real security surface, even behind Face ID or a fingerprint. It's mostly one developer shipping close to a release a day since launching in late September 2026, and the scope keeps widening (mail, crypto prices and Spotify pills are recent additions). And forks can't reuse the name, character, icon or sounds, because those assets aren't MIT.

agent-autoloop5.6k★

Agent Autoloop turns Claude Code or Codex into a background worker. A cron schedule, a new GitHub issue or an error alert becomes an event, your rules pick the agent and the folder, and the agent works, exits, and pings you only for calls it shouldn't make alone: deleting data, restarting a service, spending money. MIT licensed and free. It runs on your own Windows, Mac or Linux machine with Node 24, configured by one rules.json file. No server and no dashboard: approval cards land in desktop notifications, Feishu, DingTalk, WeCom, Slack or Discord, and you answer from there or the CLI. While it waits on you, no agent process runs and no tokens burn. Solo: free, and a good fit for unattended morning briefs, issue triage or site checks. Small teams: free, but it's one person's machine and one person's approvals. Larger teams: this is personal tooling, not shared infrastructure. To watch several agents live, superset-sh/superset or manaflow-ai/cmux fit better. The catch: the docs are Chinese-first, and the Slack and Discord templates were only tested against local mocks. The README also notes vendors steer heavy automation toward API keys, not personal subscriptions, so plan on API billing.

hapi5.4k★

Hapi runs a Claude Code, Codex, Gemini, or OpenCode session on your laptop and lets you control it from a phone or browser. The agent stays where it works best (your machine, your file system) and a web/PWA/Telegram app gives you remote eyes and a remote keyboard. End-to-end encrypted via WireGuard plus TLS. AGPL. Setup is one npx command. You get a URL and QR code, scan it, and you are looking at the active session from anywhere. Workspace browsing is opt-in. Voice control works through a built-in assistant, and Telegram approvals let you sign off on agent actions while away from the keyboard. The relay is provided, but you can self-host with Cloudflare Tunnel or Tailscale. Solo developers running long agent loops: this is the most polished way to monitor them away from the desk. Small teams: same value, but pair-sharing across teammates is not the use case. Large teams: stick to internal devloop tools. The AGPL license adds friction for enterprise legal reviews. The catch: it is built around a relay you do not own. Self-hosting is supported, but the easy path uses someone else's WireGuard hub. Read the security model before you point it at production credentials.

zeron3.7k★

Zeron is one native window for the coding agents you already use. It drives Claude Code, Codex, Cursor, Devin, Grok, Hermes, Pi and Antigravity in their own sessions instead of swapping in a new harness. The desktop app is MIT licensed, needs no account, and keeps sessions on your device. The UI is Rust on GPUI, the framework behind the Zed editor. Builds ship for macOS on Apple silicon, Windows and Linux, plus Android and iOS apps. A headless Linux install runs the engine as a service on a VPS, so agents keep working after you close the laptop. Optional sign-in syncs sessions across devices: start Claude Code at your desk, approve the diff from your phone. Solo: free, and the multi-device angle is the reason to pick it over superset-sh/superset or manaflow-ai/cmux. Small teams: free per person, but sync is per user, not a shared workspace. Larger teams: wait. A hosted Cloud option runs sessions in their own sandboxes, billed in credits with no public price list. The catch: it's young and ships releases almost daily. Devices signed in to one account can read and write each other's workspace files, so only sign in machines you trust.

openharness1.5k★

OpenHarness runs a crowd of coding agents side by side and lets you watch them all from one desktop app. Claude Code, Codex, Cursor, OpenCode, Copilot and nine more each get their own tmux pane and, optionally, their own git worktree, across any machines you link over an encrypted connection. It comes from Autonomous, the company better known for standing desks, and the app is MIT and free to download. Install is a curl script from Autonomous's own domain, and the daemon auto-updates. macOS is the main platform; Linux is in progress and Windows isn't there yet. It doesn't wrap the agents, so each one still needs its own subscription or API key. Right now you need an Autonomous account even for purely local use, and the project's own docs list removing that as an outstanding requirement. Hosted machines beyond one small free one are the paid part, with no public price list yet. Solo: worth a try if you juggle several agents. Teams: wait for account-free local mode. Claude Squad (smtg-ai/claude-squad) does the tmux-and-worktrees job with no account at all. The catch: "open source, all of it" is overstated. The sign-in service and the hosted machine manager are closed, so you can't run the whole thing yourself. And until a September 29, 2026 commit, the desktop and phone apps sent usage analytics like app opens, sign-ins and agents created.

herdrm796★

herdrm is a native macOS window onto every coding agent you have running, local or over SSH. Claude Code, Codex, Gemini, Grok, and OpenCode across your Mac and any box you can reach, each on a real PTY at full TUI fidelity, with a sidebar showing which one is working, blocked, or done. It is a client for herdr, the Apache 2.0 runtime that owns the terminals. The details show someone uses this daily. SSH targets accept user@host, explicit ports, ssh:// URIs, and ~/.ssh/config aliases, with auth falling back from OpenSSH keys to Tailscale SSH to a password kept in the macOS Keychain rather than a file. Failures name themselves instead of reporting a bare not connected. Files and images paste straight into Claude Code or Codex, locally or streamed over SSH. Free, with signed builds and a Sparkle appcast so it updates itself. The recurring cost is the agent subscriptions you already pay for. Mac only, and herdr has to be running on every machine you want in that sidebar. The catch is that there is no license file. herdr itself is Apache 2.0, but this client ships without one, which legally means no rights granted at all. Given the release cadence it is almost certainly an oversight, and it will still fail a dependency audit.

The Open Source Drop

Open source tools worth knowing about.

hello@opensourcedrop.com
ToolsExploreAlternativesLicensesScoringSuggest a ToolAboutTerms

Buying guides

AI Agent FrameworksOpen Source Agent FrameworksOpen Source Claude Code & Codex SkillsOpen Source Agent Runtimes & SandboxesOpen Source Multi-Agent SystemsPython Developer ToolsLLM Inference ToolsDevOps & Infrastructure ToolsCLI ToolsOpen Source MCP ServersAll buying guides

The Open Source Drop provides informational summaries of open source tools and their licenses. This is not legal advice. Always review the full license text on each tool's repository before using it in your project. Product names and logos mentioned on this site are trademarks of their respective owners. The Open Source Drop is not affiliated with or endorsed by any proprietary software vendor. Terms of use.