Conductor Is Joining Meta: 12 Open Source Alternatives

Key Features

• Every task runs in its own isolated Git worktree, and each worktree is a real Git worktree usable with standard git commands such as git status, git rebase, and git cherry-pick
• 27 or more coding agents supported, against the 4 that Conductor ran
• Embedded Chromium browser per worktree with Design Mode, which lets you click any UI element to send its HTML, CSS, and a cropped screenshot into an agent's prompt
• Native GitHub and Linear integration for browsing PRs, issues, and project boards in-app and opening a worktree from any task
• SSH remote execution with auto-reconnect, port forwarding, and passphrase caching, so agents can run on a remote machine
• Orchestration that lets agents manage and coordinate workspaces for larger tasks, plus an Orca CLI so agents can drive the environment via commands like orca worktree create
• MCP, hooks, and skills support, with WebGL terminal rendering, infinite splits, and scrollback restored on restart

Pros

• MIT licence permits commercial use, modification, distribution, and private use with no AGPL compliance overhead
• Cross-platform on macOS, Windows, and Linux, where Conductor was macOS-only
• Mobile companion app for iOS and Android shows live agent status and keeps terminal work moving when you are away from your desk
• Ships new features daily from a team backed by Y Combinator

Cons

• Orca is a YC-backed project. If the acquisition risk that ended Conductor concerns you, that is worth weighing, though Orca is MIT and self-run, so there is no service that can shut down on you.

License & Hosting

• License: MIT, copyright Lovecast Inc. 2026, permitting commercial use, modification, distribution, and private use
• Where it runs: local-first on macOS, Windows, and Linux, with optional SSH connection to a remote machine for running agents there
• Setup: download the desktop app from onorca.dev, bring your own agent subscriptions and API keys
• Costs: provider API key or agent subscription for each agent you run, Orca itself is free

Best For

I'd start with Orca when you want the most direct drop-in for Conductor: same worktree model, same parallel-agent pattern, broader OS and agent support, and no licence friction.

Key Features

• Fan the same prompt to several models at once, each running in its own thread and worktree, so competing attempts finish in parallel
• Side chats let you ask a question about a thread without touching it, keeping the thread working while you ask
• One button commits, pushes, and opens a pull request on GitHub with a generated title, body, and changelog
• iOS and Android apps, a web app, and an Electron desktop app all control agents running on your machine
• A subagents view follows work delegated to subagents in one place
• Inline diff review, draft PRs, stacked PRs, and amended PRs before you push

Pros

• Actively maintained by T3 Tools Inc.
• MIT licence: free for personal and commercial use, modification, and distribution
• Cross-platform via curl, PowerShell, or npx, so you can try it before committing to an install
• T3 Code does not resell tokens: plug in the credentials you already have and it orchestrates them
• Supports switching models mid-thread, with more agent integrations shipping weekly

Cons

• 6 or more supported agents (Claude Code, Codex, Antigravity, OpenCode, Cursor, Grok, and fx), fewer than the 27 or more in Orca or the 35 in Emdash

License & Hosting

• License: MIT (copyright T3 Tools Inc., 2026), permitting commercial use, modification, distribution, and private use
• Where it runs: local-first on macOS, Linux, and Windows, with mobile and web clients that control the local agent daemon
• Setup: curl -fsSL https://t3.codes/install.sh | sh on macOS and Linux, PowerShell script on Windows, or npx t3@latest to try it once without installing
• Costs: provider API key or subscription for each agent. T3 Code is free

Best For

T3 Code is the community's pick, and by T3 Tools' own account, Conductor was among the tools that inspired it. T3 Code fits a developer who wants a large community, a simple cross-platform install, and a control surface that works over their existing agent subscriptions. If you need a longer agent list or want scheduling, Orca or Emdash are the stronger picks.

Key Features

• Daemon architecture: desktop, web, mobile, and CLI clients all connect over a websocket to the same running daemon, which you can run on a laptop, a VM, or a Docker host
• 400 or more community plugins built on the same plugin API agents use, covering workspace panels, sidebar items, composer attachments, themes, and more across desktop, browser, iOS, and Android
• Native iOS and Android apps with full feature parity with the desktop app, not a companion viewer
• Schedules and heartbeats for running agents on a cron, plus a CLI that can run agents, list them, send follow-up messages, and connect to remote daemons
• Each worktree gets its own dev server URL, so 2 agents running dev servers in parallel get separate local addresses instead of port collisions
• Local speech-to-text and text-to-speech voice control, no telemetry, tracking, or forced login
• Pull requests handled in-app for GitHub, GitLab, Gitea, Forgejo, and Codeberg, where Conductor Cloud supported GitHub only

Pros

• Apache-2.0 licence is permissive and includes an express patent grant, with no commercial-use restrictions
• Multi-device access through an opt-in relay that is end-to-end encrypted and requires no account
• TypeScript SDK for automating Paseo, plus agent orchestration where agents create worktrees and launch other agents across providers

Cons

• One person maintains this, so the pace of fixes and features depends on a single maintainer, though the community plugin ecosystem adds surface area
• If you want a document-first workspace with WYSIWYG Mermaid and Excalidraw editing alongside code, Nimbalyst covers that job

License & Hosting

• License: Apache-2.0, copyright Mohamed Boudra, permitting commercial use, modification, distribution, and patent use
• Where it runs: local-first on macOS, Linux, Windows, iOS, and Android, with daemon connectable to a VM or Docker host
• Setup: download from paseo.sh/download, the daemon starts automatically when the app opens, then connect any client
• Costs: provider API key or subscription for each agent you run, Paseo itself is free

Best For

Paseo is the pick when mobile access with full feature parity matters, or when you want agents running on a VM or Docker host that you connect to from wherever you are. If you want a larger community and a team that says it ships new features daily, start with Orca.

Key Features

• Each task gets its own Git worktree and branch, so Claude Code, Codex, OpenCode, Cursor Agent, Copilot CLI, Gemini CLI, Amp, and any other CLI agent can work on several tasks at once without overwriting each other's files
• Changes pane for reviewing, committing, pushing, and opening pull requests, plus a pull request pane with checks, review threads, and merge
• Scheduled automations and event-triggered automations (issue triage, changelog drafts, dependency bumps) that run without prompting and open PRs for review
• iPhone and iPad app for answering agents and reviewing diffs remotely (requires Pro)
• CLI, TypeScript SDK, and MCP server so agents and scripts can create workspaces and launch agents programmatically
• Remote access: add any machine as a host, workspaces keep running when your laptop sleeps
• In-app browser with Design Mode and port management

Pros

• SOC 2 Type II certification with an independent audit, relevant for teams with enterprise security requirements
• MIT licence on the desktop app, same as Orca, permitting commercial use and modification
• Repos, worktrees, terminal output, and agent sessions stay on your machine, with cloud sync covering only account and organization metadata

Cons

• Linux support is experimental (AppImage only), not production-ready for Linux-first teams
• Scheduled automations, remote access, and the iPhone app require Pro at $20 per user per month

License & Hosting

• License: MIT, permitting commercial use, modification, distribution, and private use
• Where it runs: local-first on macOS, with optional remote machine hosting via Pro, Linux AppImage is experimental
• Setup: download from paradev.com, Git must be installed, Pro plan required for remote access and mobile
• Costs: provider API key or subscription for each agent, the desktop app is free, Pro is $20 per user per month, Team is $60, Ultra is $200

Best For

I'd pick Paradev for macOS developers who want remote access and mobile on Pro at $20 per user per month, or cloud workspaces on Team at $60. Skip it on Linux until the AppImage stabilises and use Orca or Pane there.

Key Features

• Main agent plans and delegates; each worker runs in its own Git worktree with its own branch
• Sessions sorted into lanes by agent, PR, CI, and review state so the most urgent item surfaces first
• Failed CI checks, review comments, and merge conflicts return automatically to the worker that owns the branch
• Pick the agent and model per task; work in structured chat or the agent's own terminal UI; each worker gets a browser to preview and test the app
• 35 or more coding agents supported, including Claude Code and Codex
• Desktop builds for macOS (Apple Silicon and Intel), Windows, and Linux (AppImage, deb, rpm)
• iPhone and Android apps paired by QR code for alerts, replies, and PR review on the go

Pros

• Apache-2.0 licence: commercial use, modification, and distribution permitted
• Local use is free and open source with no account needed; you bring your own API keys with no proxying
• The PR and CI feedback loop is built into the session lane model: workers handle CI failures on their own until a PR is approved, so you spend less time triaging build failures
• Mobile companion keeps you connected to agent runs and PR review without opening the laptop

Cons

• Collects limited usage and reliability telemetry including your GitHub username; you can turn this off in Settings
• Cloud workers that keep running after the laptop closes require a paid plan: Starter is $20/month, Pro is $40/month

License & Hosting

• License: Apache-2.0, permitting commercial use, modification, distribution, and private use
• Where it runs: local-first on macOS, Windows, and Linux, with optional paid Cloud workers that run in hosted workspaces
• Setup: download the desktop app from orchestrator.inc; the app runs the daemon, no CLI required
• Costs: local use is free; Cloud workers start at $20/month (Starter) or $40/month (Pro)

Best For

I'd use Orchestrator.inc when the work-review-approve cycle is the bottleneck: the worktree-per-task model is shared by most tools on this list, but the built-in PR lane, CI feedback loop, and reviewer agent give the clearest view of which tasks need attention and why. The telemetry is worth noting before you start; turn it off in Settings if that matters. If you do not need the PR and CI feedback loop, Orca is the more direct pick.

Key Features

• Detects and supports 35 coding agent CLIs from whatever you have installed, each running in its own Git worktree on a separate branch
• Scheduling for recurring agent runs: automate bug sweeps, release prep, reviews, and checks on a timed basis
• Pull issues from Linear, Jira, GitHub, Notion, Asana, and more and pass them directly to agents with full context attached
• SSH connection to remote dev servers, cloud VMs, or GPU boxes, with agents running remotely with full worktree isolation
• Built-in file editor for browsing, editing, searching, and recovering files directly inside the task workspace
• Library for saving reusable prompts, installing skills, and syncing MCP servers to supported agents
• In-app review flow: inspect changes, commit, push, open pull requests, and monitor CI without switching tools

Pros

• Apache-2.0 licence includes an express patent grant, same as Paseo, with no commercial-use restrictions
• Each task creates a Git worktree on its own branch by default, with an option to work directly on the checked-out branch instead
• 35 supported agent CLIs, one of the longest named lists in this roster

Cons

• Emdash is a 2026 YC-backed project with a smaller community than Orca or T3 Code
• For phone monitoring of running agents, Paseo or Paradev are the stronger picks

License & Hosting

• License: Apache-2.0, copyright General Action, Inc. 2026, permitting commercial use, modification, distribution, and patent use
• Where it runs: macOS, Windows, and Linux, with optional SSH connection to a remote machine
• Setup: download the desktop app from emdash.com, bring your own agent subscriptions and API keys
• Costs: provider API key or subscription for each agent CLI you use, Emdash is free

Best For

Scheduled recurring agent runs are the reason to pick Emdash for teams that also need Linear, Jira, or Asana tasks, or for Windows developers who want Apache-2.0 licensing and project-tracker integration in one app. If mobile monitoring matters, start with Paseo or Paradev instead.

Key Features

• Worktree sub-agents run agents in parallel on isolated branches, each in its own directory, with one-click PR creation
• Auto Run: batch-process a markdown checklist or free-text goal through agents, one task per session, with playbooks for repeat workflows
• Full CLI (maestro-cli) for headless operation, including running playbooks from cron jobs and CI/CD pipelines
• Orchestrate agents on remote hosts over SSH, so agents keep running on a machine that stays online
• Built-in web server with QR code access and optional Cloudflare tunnel for phone control from anywhere
• Pass-through to your AI provider: Claude Code, Codex, and OpenCode configurations, MCP tools, and permissions work as configured
• Automatically discovers and imports existing sessions from supported providers, including conversations from before Maestro was installed
• Real-time token usage and cost tracking per session and globally

Pros

• Cross-platform on macOS, Windows, and Linux, with keyboard shortcuts documented for each OS
• Auto Run turns a markdown checklist into a repeatable batch job without writing any glue code
• The headless CLI integrates with cron and CI/CD, so you can kick off agent runs from your build pipeline

Cons

• AGPL-3.0 licence requires publishing source code for modifications you distribute or run as a network service; check with your legal team before modifying and shipping
• 7 supported agents today, 3 of them in beta (Claude Code, Codex, OpenCode, and Factory Droid are fully supported; Copilot-CLI, Qwen3 Coder, and Oh My Pi are in beta)

License & Hosting

• License: AGPL-3.0, permitting use and modification, but requiring source disclosure for distributed or service-hosted modifications
• Where it runs: local-first on macOS, Windows, and Linux, with optional SSH connection to a remote host
• Setup: download the desktop app from the Maestro releases page; bring your own agent credentials
• Costs: provider API key or subscription for each agent you use; Maestro itself is free

Best For

I'd pick Maestro when you want repeatable batch workflows or need to run agents from a cron job or CI pipeline without a separate orchestration layer. The Auto Run task runner and headless CLI are the distinctive feature set here. If you need more than 7 agents or want a permissive licence, start with Orca (MIT) or Orchestrator.inc (Apache-2.0).

Key Features

• 8 agents in one IDE: Claude Code, Codex, Cursor, Pi, OpenCode, Grok, omp, and Hermes
• Every thread runs on its own worktree branch, with a working tree view of uncommitted changes
• One agent can spawn and manage another, each in its own thread, so a triage agent can hand sub-tasks to specialists
• Each agent uses your existing provider subscription, billed by the provider directly
• CLI open to any program: shell scripts, cron jobs, or bots can spawn a bb thread
• GitHub integration, agent memory, scheduled jobs, and remote access are all plugins, so you add only what you need
• Run anywhere with npx bb-app@latest, served at localhost:38886, without installing the desktop app

Pros

• MIT licence end to end: use, modify, and distribute freely with no AGPL compliance overhead
• Runs anywhere with npx; useful for quick experiments without committing to a full install
• Pluggable architecture: add GitHub integration or remote access only when you need them

Cons

• Desktop app is Apple Silicon first; Linux and Windows builds are alpha, so expect rough edges on those platforms
• In active development: core architecture is stable, but workflows and surfaces are still evolving
• Sends anonymous usage telemetry in production runs; opt out with BB_TELEMETRY=false

License & Hosting

• License: MIT, permitting commercial use, modification, distribution, and private use
• Where it runs: local-first on macOS (Apple Silicon); Linux x64 and Windows x64 builds are alpha; runs anywhere via npx
• Setup: download the desktop app for macOS Apple Silicon, or run npx bb-app@latest; uses the provider CLI you already have authenticated
• Costs: provider API key or subscription for each agent you use; bb is free

Best For

I'd use bb when you want an agentic IDE where agents can spawn sub-agents in separate threads, and you want a pluggable base you can extend yourself. The npx fallback also makes it easy to try without committing to a full install. If you need stable Linux or Windows support today, start with Orca or Emdash while bb's alpha builds mature.

Key Features

• Isolated workspaces in local project directories, git worktrees, or on a remote server over SSH
• Custom agent loop with Plan/Exec mode; supports Ollama for local LLMs and OpenRouter for other models
• Integrated code review with a sidebar for agent status and a git divergence view for changes and conflicts
• Cost and token tracking so you can see what each parallel run is spending
• VS Code extension for jumping into Xum workspaces directly from the editor
• Server mode with a responsive UI for mobile users
• LLM provider-agnostic; usable from the browser or desktop

Pros

• SSH remote execution lets agents keep running on a machine that stays online, without a paid cloud plan
• Multi-model and provider-agnostic: bring your own Ollama setup or an OpenRouter key with no model lock-in

Cons

• AGPL-3.0 licence requires source disclosure for modifications you distribute or run as a network service
• Runs its own agent loop rather than driving Claude Code or Codex; check that the capabilities you rely on in those CLIs are available in Xum before switching
• Pre-built binaries are available for macOS and Linux

License & Hosting

• License: AGPL-3.0, copyright Coder Technologies, permitting use and modification with source disclosure for distributed or service-hosted modifications
• Where it runs: local-first on macOS and Linux, or on a remote server over SSH
• Setup: download pre-built binaries from the GitHub releases page for macOS and Linux
• Costs: LLM provider API key or Ollama setup; Xum itself is free

Best For

I'd pick Xum when you want provider-agnostic parallel workspaces that you can point at a remote server over SSH, and you are comfortable running a custom agent loop rather than delegating to Claude Code or Codex directly. If you need Windows support or want to drive the established CLI agents with their full tool configurations, Orca or Orchestrator.inc are the stronger picks.

Key Features

• WYSIWYG editors for markdown, mockups, Mermaid, Excalidraw, CSV, data models, and code in Monaco, where both you and the agent edit the same files
• Red/green diff review: step through the agent's proposed edits in the rendered document and accept or reject each one
• Parallel agent sessions in isolated Git worktrees, with a code editor, terminal, AI git management, commits, and worktrees
• Task tracking built in: plans, bugs, features, and todos in trackers that agents read and update alongside you
• iOS companion app for seeing which agents need you, replying by text or voice, swiping through diffs, queuing the next task, and getting a push notification when an agent is waiting
• MCP server support, extensions, and an extension system to build your own visual workspace

Pros

• MIT licence, permitting commercial use, modification, distribution, and private use
• Plain files on disk: everything in git, no proprietary store to migrate out of
• Wiki plugin works in Claude Code, Codex CLI, Cowork, Cursor, VS Code, and any MCP agent with no app or account needed
• Individual app is free and works with an existing Claude Code or Codex subscription or API key

Cons

• Supports Claude Code and Codex (OpenCode and Copilot in alpha), not the broader list of Orca or Emdash

License & Hosting

• License: MIT, permitting commercial use, modification, distribution, and private use
• Where it runs: local-first on Linux, macOS, and Windows; Wiki plugin runs inside other tools without the app
• Setup: download from nimbalyst.com; app is free for individuals and works with an existing Claude Code or Codex subscription or API key
• Costs: provider subscription or API key for each agent. Individual app is free

Best For

Document-first developers who want a rendered diff review in the actual document are the right fit for Nimbalyst, especially those who use Mermaid and Excalidraw alongside code. If you need more than those agent integrations, Paseo or Orca cover those.

Jean supports 8 or more AI backends, including Claude, Codex, OpenCode, Cursor, Pi, Command Code, Grok, and Kimi Code. It uses your own CLI installations and subscriptions, with no vendor lock-in.

Key Features

• Headless jean-server for Linux amd64 and arm64 (Ubuntu 22.04+ or Debian 12+ recommended) keeps agents running after the laptop is closed, full UI accessible from any browser
• Jean says its RTK plugin cuts tool-output tokens by 60 to 90 percent
• 8 or more AI backends supported, including Claude, Codex, OpenCode, Cursor, Pi, Command Code, Grok, and Kimi Code
• Magic git commands for committing, reviewing, opening pull requests, and fixing conflicts
• GitHub and Linear integrations in a single workflow, with Linear issues loadable into any session
• Automated merging with AI fixing conflicts when merging PRs or worktrees
• Mr. Robot (beta) sweeps GitHub issues and drafts plans

Pros

• Apache 2.0 licence includes an express patent grant, permitting commercial use, modification, distribution, and patent use with no share-alike restriction
• No vendor lock-in: Jean uses your own CLI installations and subscriptions, and everything runs on machines you control
• Web access serves the full UI over HTTP and WebSocket with token authentication, so any browser including a phone connects to a running session

Cons

• Jean is tested on macOS, and Windows and Linux desktop testers are still needed, so it is not an established desktop choice on those platforms yet
• Mr. Robot issue sweeping is in beta, not a finished feature
• Smaller community than most tools in this list

License & Hosting

• License: Apache-2.0, permitting commercial use, modification, distribution, and patent use
• Where it runs: desktop app tested on macOS; headless jean-server for Linux amd64 and arm64; web UI accessible from any browser
• Setup: download from jean.build for the desktop app; for headless, download the jean-server binary for Linux and configure web access
• Costs: provider API key or subscription for each agent; Jean is free

Best For

On a Linux server that stays on after the laptop closes, Jean is the pick if you need agents working overnight or want the RTK token plugin, which Jean says cuts tool-output tokens by 60 to 90 percent. Skip Jean as a Windows or Linux desktop tool until the macOS-only testing disclaimer is resolved, and start with Pane or Emdash on those platforms instead.

Key Features

• Creates and cleans up Git worktrees automatically when a pane is created or deleted
• Each pane gets its own block of 10 ports via $PANE_PORT, so parallel dev servers never collide
• MCP server registered with Claude Code, Codex, and Cursor on launch, including inside WSL, so agents can start and check on other agents
• Sessions: chats with Claude Code, Codex, or Cursor that split up work, start panes, watch them, collect reports, and track PRs
• Remote Pane keeps the desktop UI local while worktrees, terminals, and agents run on a VM, WSL box, or another machine, self-hosted on your own hardware
• .env and gitignored files copy into every new worktree automatically
• Supports conductor.json setup scripts alongside pane.json

Pros

• Runs on Windows, macOS, and Linux, where Conductor was macOS-only
• Built-in Git: diff, commit, push, pull, rebase, squash, and merge, all from the keyboard
• Free with no paid tier

Cons

• AGPL-3.0 is free to use, modify, and distribute. Check its terms before shipping or hosting a modified version commercially

License & Hosting

• License: AGPL-3.0 with attribution requirement (copyright Dcouple Inc., 2025), free to use, modify, and distribute
• Where it runs: local-first on Windows, macOS, and Linux, including WSL; Remote Pane for self-hosted remote execution on machines you own
• Setup: download the desktop app; bring at least one supported agent CLI installed and signed in
• Costs: provider API key or subscription for each agent. Pane is free with no paid tier

Best For

Pane fits Linux and Windows developers who want a terminal-first tool and need the conductor.json compatibility as a migration aid. The AGPL-3.0 licence is worth reading before you ship or host a modified version commercially. If mobile access matters, Orca or Paseo cover that job.