An opinionated AI agent workspace with strong opinions about how AI-assisted
development should work. Git worktrees, chat sessions, terminals, and GitHub &
Linear integrations in one cohesive workflow. Run it on your laptop or a
dedicated server - connect from a browser or the native app.
Put Jean on a dedicated Linux host with
jean-server, or turn on Web Access
in the desktop app. The same workspace is available over the network -
open it in a browser, or point the native Jean app at that remote
instance and keep working with the full desktop shell.
Browser - full UI via token-protected HTTP/WebSocket (default port 3456).
Native app - title bar server icon → Add remote; switch Local ↔ remote anytime.
Tailscale recommended - private mesh access instead of exposing a public port.
Desktop, browser, phone - same projects & sessions
Server
jean-server --host tailscale
Web
https://host:3456?token=…
Native
Add remote → connect
Most useful features
The parts of Jean you'll use every day.
Linked Projects, Isolated Workspaces, Magic Git Commands, Magic Prompts,
Mr. Robot, and server/web access are the high-leverage workflows that make
Jean more than a chat window for agents.
Linked Projects
Make multiple repos feel like one knowledge base.
Link related projects once and Jean makes your AI agents aware of the
surrounding codebase. Your app repo, docs repo, backend, SDK, and
shared libraries can all contribute context to the same session -
without copy-pasting files or rebuilding prompts by hand.
Agents are reminded to check linked repos for CLAUDE.md, AGENTS.md, and docs/.
Saved contexts from linked projects appear together, so prior decisions travel across repo boundaries.
Linked projects are available as shared context, so agents can understand and coordinate changes across related repos.
Agent Workflows
Parallel agents, magic commands, and automation.
Jean gives every agent a safe workspace, turns repetitive repo work into
guided AI workflows, and can even sweep GitHub issues for you in beta.
Isolated Workspaces
Run agents in parallel without chaos.
Every task gets its own git worktree, so agents can investigate, edit,
test, and propose changes independently while your main repo stays safe.
•Launch multiple agents at once without branch conflicts or file clobbering.
•Review each worktree separately before merging anything back.
•Archive, restore, or delete agent workspaces from Jean instead of juggling shell commands.
Magic Git Commands
Ship Git work without the busywork.
Investigate issues and PRs, run AI code reviews with tracked findings,
generate commits and PR descriptions, write release notes, and resolve
merge conflicts from one command palette.
•Load GitHub and Linear context directly into an agent session.
•Review diffs, track findings, and turn agent output into actionable fixes.
•Create commits, PR copy, release notes, and conflict resolutions with AI help.
Every built-in workflow has an editable prompt template plus per-prompt
backend, model, and reasoning controls. Use fast models for routine
copy and stronger reasoning for reviews or investigations.
•Customize the exact instructions behind commits, reviews, PRs, and investigations.
•Choose different backends, models, and effort levels per workflow.
•Keep the UI stable while changing how the AI thinks and writes.
Mr. Robot polls open GitHub issues, creates one Jean worktree per issue,
and drafts a focused plan. If you want, it can yolo approved plans too.
•Schedule issue sweeps with limits for issue count and active worktrees.
•Filter by included or excluded GitHub labels to control what gets picked up.
•Choose separate planning and yolo execution backends and models.
Everything you need.
Opinionated by design. We picked the best workflows so you can focus on shipping.
Worktree Management
Automated worktree creation, archiving, and restoration. Manage
branches and workspaces without touching the command line.
Automated Merging
Merge GitHub PRs or local worktrees automatically. Built-in merge
conflict resolution powered by AI.
Token-Reduction Plugins
One-click installs in the Opinionated pane. RTK proxies your CLI
tools to cut 60-90% of tool-output tokens across every backend;
Caveman trims Claude's replies by ~65-75% with terse, technical
output.
Run Jean on a Linux server with jean-server,
or enable Web Access on the desktop app. Connect from any browser
or from the native app (Add remote). Prefer Tailscale for private network access.
Free forever under the Apache 2.0 license. Inspect the code,
contribute, and make it your own.
Opinionated by Design
No endless configuration. Jean has strong opinions about git
workflows, workspace isolation, and AI orchestration - so you
spend less time configuring and more time shipping.
Linear Integration
Investigate Linear issues, load context into sessions, and configure
per-project API keys and team settings. Your Linear workflow, built in.
Want to learn more?
Dive deeper into Jean's workflows, integrations, and configuration.
Our docs cover everything from your first session to advanced setups.
Jean is an open-source AI assistant that helps you manage multiple
projects, worktrees, and sessions with Claude CLI, Codex CLI, OpenCode, Cursor, Pi, Command Code, Grok, and Kimi Code.
Use it as a desktop app, or run it on a server and connect from a
browser or the native app - same projects and sessions either way.
Yes. Install jean-server on Linux
(or enable Web Access in the desktop app). Clients connect over
token-protected HTTP/WebSocket: open the URL in a browser, or in the
native Jean app use the title bar server icon → Add remote.
We recommend binding on Tailscale (or similar) instead of the public internet.
See
Web Access and headless.
Yes, 100%. Jean is open source under the Apache 2.0 license. There
are no paid tiers, no feature gates, and no usage limits. You bring
your own subscription for any supported AI coding agent. If you'd like to support development,
consider sponsoring us.
The desktop app is tested on macOS; Windows and Linux desktop builds
work but need more testing. Headless jean-server
targets Linux amd64/arm64. Download release builds from GitHub - Node.js
and Rust are only required if you build from source.
Jean uses git worktrees to create isolated workspaces for each AI
agent session. It automates common git operations like reviewing code,
creating commits, opening PRs, and resolving merge conflicts. Each
workspace is independent, so multiple agents can work on different
branches simultaneously.
Jean supports Claude CLI, Codex CLI, OpenCode, Cursor, Pi, Command Code, Grok, and Kimi Code. Pi, Command Code, and Kimi Code are currently beta integrations. You bring
your own subscription or API access for each provider.
Yes. The Opinionated pane installs RTK (a CLI proxy that compresses
tool outputs 60-90% across every backend) and Caveman (a Claude
skill that produces terse replies, ~65-75% shorter) with one
click. See
Opinionated Plugins.
We welcome contributions of all kinds - code, documentation, bug
reports, and feature requests. Check out the
repository's CONTRIBUTING.md
to get started.
We believe developer tools should make decisions for you, not present
you with a wall of configuration options. Jean picks the best
workflows - git worktrees for isolation, automated merging,
structured context loading - so you can focus on building. If a
workflow is clearly better, we just do it that way.