Give every task its own git worktree and agent. Watch them all from one sidebar, then review the diff and ship. Each task gets its own worktree and agent. Watch them all, then review and ship.
Watch
Every card shows whether its agent is working, done or waiting on you. When one asks a question you hear it, and one shortcut takes you there.
Find
The command center opens on the agents that need you. Or type a word from a workspace's title, branch or description to find it among the rest.
Preview
The globe gathers each workspace's dev servers, desktops, links and files. Open any of them beside the agent, and pages reload as it keeps working.
Ship
When an agent opens a pull request, its card shows the checks running. Move on to other work: a badge lands on the card when they pass.
Customize
Describe it, like one that deploys the branch for preview. You approve every command it runs, and the card tells you when the deploy is live.
Start
Type or dictate what you want, pick any agent CLI, and it starts in a workspace of its own while you keep working.
Review
Roll back a block the agent got wrong or edit a line in place, in a split or unified view.
Komla runs each agent's own CLI with its own subscription or API key, so there are no keys to add. Komla doesn't charge for tokens.
Every feature, unlimited agents and unlimited local workspaces from the first launch. A license is only for running more than 4 SSH workspaces at once.
The source is public on GitHub under the AGPL, so you can see exactly what the app does on your machine.
You need Git and an agent CLI, such as Claude Code or Codex.
Connect to any machine over SSH. Your agents run there instead of on your laptop. Up to 4 SSH workspaces can run at once, and a license lifts the limit.
Run any agent CLI you want in each pane, next to your tests or a dev server.
Give it a mood or a few colors. Komla previews the theme across the whole window and keeps it only if you apply it.
With the komla CLI or MCP server, an agent can split a task across new workspaces, each with an agent of its own.
I'm deep in the weeds in the AI world and I've tried so many different orchestrators, and I've not resonated with a single one until Komla. I typically sling 10+ terminal sessions across 4-5+ projects — I'm now moving all that to Komla. It's simple, takes custom start commands for the agents, and it just works.
I love the product and have been immersed in it since I first bought the license. Keep up the good work, you're on to something great here!
The app you've built has been tremendously helpful with improving my workflow.
Run a desktop app on a Linux host with komla desktop run, and use it
over VNC or in a browser tab.
Agents add their own items with komla status set, or you define items
backed by commands, with icons, colors, links and popovers.
Edit Markdown as rich text that saves back only what you changed, and draft prompts in per-workspace notes kept out of the repo.
See each workspace's pull request, CI checks and tags at a glance.
Run an agent on a schedule, each run in its own workspace. On an SSH host, runs happen even with every device off.
Check on agents and answer them from your phone, tablet or another computer over your own Tailscale network, with a push notification when one needs you.
Komla finds the dev servers in your terminals and opens them beside the agent, tunnelling ports from SSH hosts for you.
Double-tap ⌘ or Ctrl to dictate a prompt. Runs offline in 25 languages, so audio never leaves your machine.
Fetch, pull, rebase and push from the git menu, compare branches, then open a pull request on GitHub or GitLab.
Save shell commands and agent prompts, then run them in any workspace from the tab bar.
⌘P or Ctrl+Shift+P searches every workspace, command and shortcut, and finds files by name or content with ripgrep.
Describe the look you want and an agent builds the theme, or pick one like Catppuccin or Dracula.
Add a VM or container as an SSH host, and let agents skip permission prompts away from your own files.
Agents create workspaces, start other agents, open previews and change Komla's
settings once you confirm, with the komla command in every terminal.
An MCP server is there too.
Every workspace across your projects in one inbox, with the ones that need you first and what each agent asked.
See when an agent finishes, fails or asks for input, with optional sounds.
komla notify does the same for any script.
Run your own commands when an agent finishes, needs input, or a workspace starts, on this computer or the SSH host.
Presets for Claude Code, Codex, Cursor CLI, Copilot CLI, Amp, Qwen and more. Any other agent works as a custom command.
Copy files, install dependencies and run commands when a workspace is created. Your agent can read the repo and propose the steps.
Sync an SSH workspace to this computer to run desktop and mobile apps locally.
Link workspaces from different repos so one agent can read and edit across them.
Browse a workspace's commits as a graph and see the changes in each one.
See who last changed each line, and click through to that commit.
See every commit that changed a file, including across renames.
Browse and edit files in a built-in Monaco editor, with optional Vim mode. Open files update as the agent rewrites them.
Terminals, browsers and notes in tabs and split panes, arranged by drag and drop.
Archive, stop or delete workspaces in bulk, sort by activity, and filter by project or host.
Every settings change, made in Komla, by an agent or by hand, is listed and can be reverted.
Open several windows and move workspaces between them.
Change any shortcut, or reset them to the defaults.
Yes. Every feature is free, with no account and no trial. You can run unlimited agents in as many local workspaces as you like, and in up to 4 SSH workspaces at once. You only need a license if you want more than 4 SSH workspaces running at the same time.
Yes. Komla's source code is public on GitHub under the GNU Affero General Public License v3.0 (AGPL). The downloads on this site are the official builds, which update automatically and come with the built-in AI helpers ready to use.
Yes. Komla runs each agent's own CLI, so it uses the login or API key that CLI already has. Komla doesn't charge for tokens.
Your code stays on your computer unless you generate a commit message. Komla has two optional AI helpers: one names new workspaces from your prompt (on by default) and can name branches too (off by default), and one writes a commit message from your staged diff when you click generate. That text goes through OpenRouter to Google, OpenAI or Mistral, which don't train on it. Komla pays for it and doesn't store it. You can use your own key or turn the helpers off in settings. The agents you run (Claude Code, Codex and others) talk to their own APIs, like they would in any terminal.
Any agent that runs in a terminal. Claude Code, Codex CLI, OpenCode and Antigravity CLI come with one-click notifications and the Komla skill. Komla also has presets for Gemini CLI, Cursor CLI, GitHub Copilot CLI, Amp, Qwen Code, Mistral Vibe, Goose and more. Any other agent can set up its own notifications with a self-integration skill, or you can add it as a custom command.
Yes. Each workspace is a normal folder on disk. Open it in VS Code, Cursor, Windsurf or Xcode with one click, or point any other editor at it.
Archive a workspace to hide it, or delete it to remove the worktree and branch from disk. They're standard git worktrees, so you can also manage them with git.
Yes. Komla uses your local Git to create and manage worktrees.
