Local development,
without the duct tape.
Orca is a free, open source desktop app for containers, compose stacks and Kubernetes. It gives every service an HTTPS hostname, runs a full copy of your app for each git worktree, and lets your AI agents drive it all. One file in your repo sets it up for the whole team.

One app instead of Docker Desktop, a reverse proxy you keep alive by hand, and a wiki page of setup steps.
Orca in four minutes
Recorded from the real app: a project folder that just runs, every container live, a whole team set up from one repo, and Claude setting up and fixing apps through Orca.
The video plays from YouTube once you start it. Watch it on YouTube
From git clone to HTTPS in a minute
Orca finds every compose project on your machine, however you start it. Add one file and it gets real hostnames, setup commands and links for everyone on the team.
Start your project
Open the folder in Orca, drop it on the window, or start it the way you always do. Orca picks it up as a stack.
git clone git@github.com:acme/shop.git
cd shop && docker compose up -dDescribe it once
Commit an orca.yaml next to the compose file, or let Orca write it from what runs.
gateway:
- hostname: shop
service: web
port: 3000Open it
Trusted HTTPS on a real name, with WebSockets and hot reload working as they should.
Any project folder, one click from running
Point Orca at a project and it shows what's inside before it starts anything: the compose file, its services and profiles, the orca.yaml, and whether it's a git repository you can run per worktree.
- Open a folder, or drop it on Orca's window. No compose file yet? Orca writes a first one
- Jump to the code: open any stack in VS Code, Cursor or your terminal, or a shell in any service
- Press ⌘K to find any stack, container, image or URL and run any action


A new developer, up and running in minutes
One repo describes your team's whole local setup: the shared services, the network, the gateway on your own domain with its certificate, and every project. Join it once, and any project is one click away, or one sentence in Claude: "spin up billing".
- Clones the repo, creates the
.env, runs the setup command and starts it, with progress you can follow - Secrets come from Passbolt or 1Password: a direct link to each entry, or filled in automatically by their CLI
- Coming from a devproxy-style HAProxy? Import its routes; projects don't change

One database server for all your projects
Instead of a MySQL or Postgres container in every project, a project says which database it needs in its orca.yaml, and gets a database and a user of its own on one server per engine on your machine. A workspace sets those servers up the same way for everyone on the team; without one, Orca runs them itself.
- MySQL, Percona Server, MariaDB or PostgreSQL, in the versions you run in production
- The connection goes into the project's
.env; its own database container stays down - Connect, snapshot, restore or reset any database from the Databases page
- Moving over? Import copies a project's data across, or ask Claude to move them all

Every service gets a real HTTPS name
A managed gateway routes shop.localhost, api.localhost or your team's own domain to the right container, with certificates from Orca's own certificate authority.
- Routes from
orca.yaml, container labels, existing Traefik labels or one click - Path routes, WebSockets, SSE and HTTP/2 on the same host
- The same names resolve inside containers, so OIDC and server-to-server calls just work
- A live landing page lists every service with its status, and its search box makes it a great start page

A running copy of your app for every branch
Work on three branches at once, or let three AI agents do it. Each git worktree runs as its own instance with its own URLs and ports, while databases and other infrastructure run once and are shared.
- Hostnames per instance, like
web-feature-x-shop.localhost - Shared data, a database of its own, or a private server, per service
orca upfrom the worktree;eval $(orca env)for tests- Idle instances stop themselves and wake on the next request

One file sets up the project for everyone
Routes, links to staging and production, build targets and setup commands live in the repo. Edit it as a form with the stack's real services and ports to pick from, or as YAML. Comments survive either way.
- Applied the moment it changes, however the stack was started
orca.local.yamlfor what differs per machine- Compose profiles switch optional services on and off, per developer

gateway:
- hostname: shop # https://shop.localhost
service: web
port: 3000
- hostname: shop
service: api
port: 8080
path: /api/* # same host, another service
hooks:
after_up:
- "api: php artisan migrate --force"Your AI runs Orca. One click to connect.
Connect Claude Code, Claude Desktop, Cursor, VS Code, Windsurf or Codex from Orca's settings. Then just ask: start a project from a folder, switch on a profile, find out why a container keeps restarting, or run a branch as its own instance and test against it.
- No tokens to copy: the AI tool runs
orca mcp, which signs in by itself - 74 tools: workspace projects, stacks, profiles, gateway routes, builds, Kubernetes, worktrees
- Works with the AI subscription you already have. No API key needed
The rest of your container workflow, too
Orca is a complete replacement for the desktop container tools you use today.
Containers
Run, inspect, edit limits live, open a terminal, browse files and follow logs with search.
Learn moreCompose stacks
Every compose project, however it was started, with profiles, editing and setup hooks.
Learn moreImages and scans
Pull with tag search, build with live output, and scan for CVEs with Trivy.
Learn moreKubernetes
A local k3s cluster in one click, with Helm, topology and a YAML editor, on its own context.
Learn moreApp Catalog
Databases, mail, monitoring and more in one click, with generated secrets and setup guides.
Learn moreRemote hosts
Install the daemon on a Linux server and manage your whole fleet from one window.
Learn morePush to deploy
A webhook from GitHub or Docker Hub pulls the new image and redeploys, keeping your config.
Learn moreCLI and API
Everything is scriptable: the orca command, a REST API and an event stream.
Learn moreNative on every desktop
Orca sets up what each platform needs on first start. No Docker Desktop required.
macOS
A lightweight Linux VM with Lima and Apple's Virtualization framework, VirtioFS mounts that keep file permissions, and ports forwarded to your Mac.
Your platformWindows
Docker inside WSL2, installed and configured for you, with memory, CPU and swap settings in the app.
Your platformLinux
Talks to Docker or Podman directly, with no VM in between. AppImage and .deb packages.
Your platformcurl -fsSL https://orca-desktop.com/install-daemon.sh | sudo shOpen source
MIT licensed. Read the code, file issues, send pull requests.
No telemetry
No analytics, no tracking. Orca only talks to the services you point it at.
Light
A Rust daemon and a native Tauri window, not a bundled browser.
Local first
The desktop daemon listens on localhost, and every request needs your token.
Is Orca free for my company?
Yes. Orca is MIT licensed, for personal and commercial use, with no account and no seat limits.
Do I still need Docker Desktop?
No. On macOS Orca runs Docker in its own Lima VM, on Windows inside WSL2, and on Linux it uses Docker or Podman directly. Already on Docker Desktop? Orca brings its images, volumes and compose projects over once: see Moving from Docker Desktop.
Does it work with the compose files we already have?
Yes. Any compose project you start shows up as a stack, and an orca.yaml is optional. Orca doesn't change your compose files; worktree instances use a generated override file instead.
What does Orca send over the network?
No analytics or tracking. Orca checks GitHub for updates, pulls from the registries you use, and talks to an AI provider only if you set one up. Security and privacy lists the details.
Which Macs are supported?
The macOS build is for Apple Silicon.
Can my AI agent use it?
Yes. Orca is an MCP server. Connect Claude Code, Claude Desktop, Cursor, VS Code, Windsurf or Codex with one click in Settings, without copying a token. Agents and MCP shows how.
Do I need an AI API key?
No. Connected over MCP, Claude and the others use the subscription you already have. Orca's optional built-in assistant runs on a local model from Ollama, LM Studio or llama.cpp, for free; an API key is only needed if you want it to use a cloud model. See Built-in assistant.
Set it up in five minutes
Download Orca, open a project, and give it a hostname. The quickstart walks you through it.