Describe what you want.
Your AI team builds it.
Orkka gives you a named team of AI agents — a project manager, developers, a tester, a reviewer. Tell them what you want, approve their plan, and they write the code, test it, and show you exactly what changed before anything goes live. Works on an app Orkka creates and hosts for you, or on code you already have.
Powered byThe Problem
You notice something wrong with your app while you're out. You write yourself a note and hope you remember. Later you sit down, find the problem again, work out what to change, and finally fix it — or wait for someone who can. The fix was small; the process ate your afternoon.
The Orkka Way
You notice the problem, open Orkka on your phone, describe it in 30 seconds and add a screenshot. Luna turns it into a clear task, a developer agent makes the change in its own private copy of the code, Scout tries it out in a real browser, and Sentinel gives it a second look. You get a plain summary of what changed, tap approve, and it's done.
Meet your AI development team
Every project gets its own team of AI agents, each with a name and a job. Rename them, give them instructions, and choose which AI model powers each one — Claude, GPT, Gemini, and more.
Luna
PMTurns your request into a clear task, asks the right questions, and writes down what “done” looks like before anyone starts.
› I've refined your task and identified 3 acceptance criteria.
Nova
Full Stack DevBuilds whole features end to end — what people see on screen, what runs behind it, and the data it keeps.
› Wired the new endpoint into the dashboard. Tests are green.
Archie
Solution ArchitectKeeps a living diagram of how your system fits together, and advises on design choices, trade-offs, and growth — one chat away.
› Sketched the event flow — want a queue between API and worker?
Pixel
Frontend DevBuilds the screens people see and tap — polished on phone and desktop, in whatever frontend technology your project uses.
› Implementing the dashboard with responsive grid layout...
Atlas
Backend DevBuilds the behind-the-scenes machinery — data storage, business rules, and the connections between them. Detects your stack automatically.
› Created REST endpoints with proper auth middleware.
Nomad
Mobile DevBuilds Android and iOS apps — native, or cross-platform with React Native, Flutter, and MAUI — matched to your project.
› Compose screen is done. Offline cache and tests included.
Rune
Game DevWrites gameplay systems in your engine's idioms — Unity, Godot, Unreal, Phaser — and keeps everything running smoothly.
› Pooled the projectiles — the spawn stutter is gone.
Muse
DesignerCreates logos, icons, illustrations, and app graphics. Everything lands in the Design gallery for you to approve or iterate.
› Three logo concepts are ready in the gallery.
Sprite
Game DesignerDraws sprites, tilesets, item icons, and HUD elements. Assets land in the Design gallery, ready for the dev agents to wire up.
› Six enemy sprites are in the gallery for review.
Scout
QAClicks through your live app in a real browser, checks every requirement, and writes automated tests so it stays working.
› All 12 test cases passing. Coverage at 94%.
Sentinel
Code ReviewerGives every change a second look before it's accepted — catches mistakes and keeps the code tidy, with comments line by line.
› Found 2 issues in the PR. Requesting changes.
Forge
DevOpsHandles the plumbing that builds, packages, and publishes your app — CI/CD pipelines, Docker, Helm, and infrastructure as code.
› Docker build succeeded. Deploying to staging...
Sage
DBAKeeps your database healthy — reviews how data is structured and looked up. Advisory only: Sage never touches a live database.
› That query needs a composite index on (org_id, created_at).
Lex
LegalDrafts terms of use and privacy policies, and flags compliance questions early. Always defers to licensed counsel for the final word.
› Drafted your privacy policy — the GDPR section needs your review.
From your first request to a finished update
Start from nothing or bring code you already have — either way, the same steps carry every task from request to done, one at a time or hands-free.
Start with any project
Let Orkka create and host a brand-new app for you, or connect code you already have on GitHub — any technology. Your AI team is ready instantly, with the right developer matched to each part of the project.
Plan the whole thing — or just one task
Describe what you want to build and Orkka drafts a roadmap: milestones and tasks in order, which you can edit. Or write a single task and let Luna turn it into a clear checklist of what “done” means. Nothing gets built until you approve the plan.
Agents build, test, and review
Each task goes through the same steps: plan, build in a private copy of the code, test in a real browser, and a second-opinion review. Run it hands-free with Autopilot or work alongside an agent in a live session — and with a runner on your own computer it all costs $0.
You have the final say
You stay the decision-maker: approve plans, answer questions, see every change explained, and accept it when you're happy — from your desk or your phone, without leaving Orkka. For developers, that is a real pull request, reviewed line by line and merged in-app.
Watch a task move across the board
Every task moves through the same columns, from idea to done. The agents hand the work to each other — you step in only for the big decisions.
Set the roadmap. Flip on Autopilot.
Plan the big milestones with your PM, put them in order, and let the team work through the board hands-free — plans approved, questions answered, reviews resolved. You come back to finished work, not a list of things waiting on you.
Each milestone becomes tasks in the right order — the behind-the-scenes part is built before the screen that needs it.
- Approves plans and answers agent questions with the recommended option — no 3am questions
- Works through milestones in order, and never starts a task before the ones it depends on
- Every change still gets reviewed — anything found goes straight back to the developer
- Stops at your spend cap and parks anything that truly needs a human
Pair-program in a live session
Open a session and chat with an agent working on a real branch, in a real worktree — like a Claude chat that ships code. Watch every edit and command stream in, steer mid-task, and approve anything risky from the browser.
- A real branch and isolated worktree from the first message
- Commands that reach outside the worktree — like git push — wait for your click
- Finish the session and it commits, pushes, and hands the branch to code review
- Ended with nothing to ship? It closes clean — no empty PRs
Added the field with inline validation and a /v1/promo/validate endpoint. Tests pass — want me to push?
git push origin feat/checkout-flow
Watch the code as it's written
The board becomes a workbench. Pin a task, keep its chat beside you, and see every file the agent touches appear in a live tree — with the exact diff, syntax-coloured, the moment it lands. Need a change? Say it in the chat; the agent picks it up mid-run.
Validate promo codes at checkout
No files changed yet.
Select a file to see its changes
A file tree that builds itself
Every file the agent edits shows up the instant the tool call streams — added, modified, removed, with line counts.
Exact diffs, not guesses
The worktree is re-diffed after each change on the machine running it, so generated files, migrations and lockfiles show too.
Steer from the chat
Ask for changes on the fly. The conversation sits right under the task, and the agent reads it while it works.
Readable code everywhere
One syntax theme across diffs, files, plans and messages — keywords, strings, types and numbers you can tell apart.
A team that gets smarter every task
Most agent tools forget everything when the run ends. Orkka’s team keeps a shared memory, follows the playbooks you write, and runs standing orders on a schedule — so week ten is faster than week one.
Shared project memory
Decisions, conventions and lessons live in one place agents and humans both read. Every run retrieves what is relevant — pinned rules first — and agents save what they learn with a remember tool. You can read, edit, pin or delete any of it.
Skills & playbooks
Write a SKILL.md — groom the backlog, cut release notes, run a security pass — assign it to agents, and they read it only when the work matches. Type /skill in chat to run it as a playbook.
Automation & autonomy
Schedule playbooks, open and close autopilot windows with a spend cap, let agents set themselves reminders, and turn GitHub issues or failed CI runs into tasks — behind one Autonomy switch you can flip off any time.
See it in action
Where your AI team works — the live board, each task's conversation, changes to review, and a living map of your system.
Everything you need to build faster
Run your AI development team from anywhere — every step happens inside Orkka, on desktop or on your phone.
Live task board
Tasks move across columns in real time as agents plan, build, and test. Skip steps or reorder columns to match how you work.
Code View
Pin a task and watch the code it produces: a live tree of every file the agent touches, exact syntax-coloured diffs, and the chat right there to steer mid-run.
A conversation per task
Every task is a conversation: agent messages, plans, questions, and attachments in one place. Approve plans or answer questions from your phone.
Review changes in-app
See exactly what changed, with Sentinel's comments line by line, then approve and merge — no need to open GitHub.
Browse your code
Connect your repositories through the GitHub App and browse files, branches, and changes right inside Orkka.
AI-Written Documents
Agents write and maintain project docs with full revision history. Chat with your PM to create or update them.
Design Gallery
Muse turns requests into logos, icons, and illustrations. Approve, reject, or iterate on each asset with one click.
Architecture Board
Archie keeps a living diagram of your system on an interactive board — ask for changes in chat and watch it update.
Notifications & Live Activity
Get notified when an agent replies, a plan needs approval, or a task gets blocked. Everything updates in real time.
Pay for what you use
Pay only for the AI work done in the cloud, with 10,000 free credits to start. Costs tracked per task, agent, and project — with budget alerts and auto-pause.
Autopilot
Flip one switch and the team works the roadmap hands-free — plans auto-approved, questions auto-answered, reviews resolved — inside the spend cap you set.
Roadmap & Milestones
Sketch milestones with your PM and have each one broken into tasks in the right order. The team works them the way you meant.
Interactive Sessions
Chat with an agent on a live branch — watch edits stream in, approve risky commands, and finish straight into code review.
Team stats & activity timeline
Per-agent runs, success rate, duration and cost on the Team page, and a Gantt-style timeline of every execution — the morning-after-autopilot screen.
MCP, API keys & webhooks
Drive Orkka from Claude Code, Cursor or scripts through the built-in MCP server and REST API with a scoped key; get signed JSON webhooks for any event, straight into n8n, Zapier or your own service.
Tamper-evident audit log
Every management action and agent run, with who did it, in a per-organization hash chain you can verify and export — the trail compliance asks for.
Have an idea? Describe it. It goes live.
No code, no repositories, no servers. Tell your AI team what you want to build, and Orkka creates the app, hosts it, and keeps improving it from the same board everyone else uses.
It starts as a conversation
Your PM asks what the app is for, who uses it, and whether people need to sign in. Then the team builds the first version.
Every hosted app comes with
- Sign-in and accounts, when your app needs them
- A database and file storage — already wired up
- AI features through Orkka, billed to your account — no API keys to manage
- A live address from day one, updated with every finished task
Hosting and database are covered by a flat monthly fee per app. Agent work is billed like any other Orkka task — on Orkka's computers, or free on your own machine with your AI subscription.
Nothing technical to manage
No repositories, branches or deploys to think about. You see tasks, reviews and a live app.
The same AI team
Luna refines, Nova builds, Scout tests, Sentinel reviews — on Orkka's computers, or on yours with your own AI subscription.
Yours to grow
Every finished task goes live on its own. Ask for the next feature from your phone; wake up to it running.
Run Orkka on your own machine
Orkka Local is the free, self-hosted edition. The same AI dev team and pipeline — working on your local repos, powered by the AI subscriptions you already have: Claude, Codex (ChatGPT), Gemini, Grok, or GitHub Copilot. No credit card, no API keys, no code leaving your machine.
How it works
A lightweight Docker core runs the board, API, and database. The Orkka runner — one small binary on your machine — does the real work: it runs your provider's CLI (claude, codex, gemini, grok, copilot, or the open-source opencode) under your subscription login and manages git worktrees over your local repositories.
- Runs on your existing AI subscriptions — Claude, ChatGPT, Gemini, Grok, or Copilot — no API keys, no per-token billing
- Your code never leaves your machine
- Agents build and test with your locally installed toolchains
Quick start
From zero to your AI dev team in three steps.
- 1Install the prerequisites
Docker Desktop and the CLI of your favorite provider — Claude Code, Codex, Gemini, Grok, Copilot, or OpenCode.
- 2Log in to your provider
Agents run on your subscription, exactly like using the CLI yourself.
$ claude /login - 3Start Orkka Local
Download the orkka-local binary (a single file, no dependencies) and run one command. It pulls the images, starts everything, and opens the app.
$ orkka-local up
A real terminal, one click away
Pop open a full terminal from anywhere in the app — running directly inside your repository, on your machine. git, npm, dotnet, vim: whatever you have installed just works, colors and all.
- Opens in your repo’s working copy — no cd, no setup
- Full PTY: colors, arrow keys, Ctrl+C, even vim and interactive prompts
- Floats on every page and keeps its session while you navigate
Private by design
Repos, diffs, and agent output stay on your hardware. The stack binds to 127.0.0.1 only — nothing is exposed to the network.
Your machine, your rules
Agents work in isolated git worktrees and open pull requests you review inside Orkka. Merges land straight into your local repository.
The same pipeline
Board, planning, development, code review, documents, architect — the full Orkka experience, self-hosted.
Your AI team, in your pocket
Orkka is coming to your phone. Follow your team's progress, approve finished work and ask for the next feature from wherever you are.
The same board, on the go
Everything happening in your project reaches your phone the moment it happens. Nothing to babysit at a desk.
- A push notification the moment a task needs you
- Review and approve finished work with one tap
- Start the next task by chatting with your team
Android ships first, iOS right behind. Follow us to hear the moment it lands: X · Instagram
"Add a page with opening hours" is ready for your review.
Free on your machine. Pay only for cloud work.
Running on your own machine with your own AI subscription is free — always. You pay when Orkka's computers do the work, or when you grow past one project.
What you actually pay for
Free for one project. $10.00/month for unlimited projects and repos.
Free and unlimited on your own machine. In our cloud you pay per token — 10,000 credits to start, no card.
Only if Orkka hosts the app for you: free for 7 days, then $10.00/month per app.
These are three separate meters. Running Orkka on your own machine, on your own AI subscription, costs nothing at all — and most people never pay for more than one of the three.
No credit card required
- 1 project, 2 code repositories, 1 person
- Unlimited agent runs on your machine — your Claude, GPT, Gemini, Grok, or Copilot subscription
- Everything: board, planning, testing, review, docs, architect
- 10,000 one-time cloud credits to try cloud execution — no credit card
ProFor solo buildersFlat. No seats, no per-project math.
- Unlimited projects and repos — for one person
- Everything in Free — local runs stay free and unlimited
- Ship from your phone: create tasks and run the whole pipeline from anywhere
- Cloud runs stay optional and separate — prepaid credits, only when our infrastructure does the work
- 5,000 cloud credits on the house, once, when you subscribe
Pro turns your orca gold — it swims in your top bar for as long as you are on the plan.
Charged per app, not per account. Starts only when you add a card.
- Your app live at yourname.orkka.app, with a database and hosting handled for you
- Free for 7 days at its own address — no card, delete it any time
- Add a card to keep it online — without one, the app pauses when the trial ends and is kept for 30 days
- 2,000 credits of agent work included every month, per app
- Does not use up your plan's project allowance
Bring your own repo instead and you never pay this — you host it yourself.
Cloud runs — optional, metered honestly
Only work executed on Orkka's infrastructure is ever billed.
- Provider cost + 10% — vendor intro prices passed straight through
- Prepaid credits with live cost tracking per task, agent, and project
- Budget caps with alerts, and auto-pause when a cap is reached
- Claude, GPT, and Gemini models — the same pay-per-token rule for all
Running on your own computer? Tasks go there and cost $0 — the cloud is the fallback, never the default.
Example rates
USD per 1M tokens, 10% platform fee included. Every model and rate is in the app's model picker.
Examples, not the full catalog. Provider base rates may change; changes are passed through.
Your subscription or ours
On your machine, agents run on the AI subscriptions you already pay for. In the cloud, Orkka's keys do the work. Nothing to configure either way — no API keys, ever.
Real-Time Cost Tracking
See costs per task, per agent, per project. Usage charts and daily spend breakdowns.
Budget Controls
Set monthly limits per project or org-wide. Get alerts before you hit your cap.