Klaussy
From first prompt to merged PR, with agents that know your codebase.
Trusted by engineers at
Source-available under SUL 1.0. Free for personal use and development.
Individual developers pay nothing. You only need a license to ship it commercially or redistribute it.
Need a different platform? All downloads ↓
- macOS · Apple Silicon M1 / M2 / M3 / M4 Macs (2020 and later) .dmg
- macOS · Intel Pre-2020 MacBooks, iMacs, Mac mini .dmg
- Windows · Installer Standard install on Windows 10 / 11 .exe
- Windows · Portable Single-file .exe, no install needed .exe
- Linux · Debian / Ubuntu Ubuntu, Debian, Mint, Pop!_OS .deb
- Linux · AppImage Works on any Linux (Fedora, Arch, openSUSE, …) .AppImage
Not sure if your Mac is Apple Silicon or Intel? Click the Apple menu → About This Mac. If the Chip row says “Apple M1/M2/M3/M4” you have Apple Silicon. If it says “Intel”, pick the Intel build.
The desktop is the face. The klaussy-agents engine is the spine. Other orchestrators stop at a worktree grid. The open-source engine underneath learns your repo's conventions and holds agents to them with custom skills and git-level guardrails.
No server. No markup. Use any agent.
Klaussy is a desktop app, not a platform you log into. There's no account, no backend, and no business model that depends on your code passing through us.
Serverless by design
Everything runs on your machine. We don't run a backend, so there's no copy of your repos, prompts, or history on someone else's disk.
no account · no cloud sync · no backend
Bring your own subscription
Not a token reseller. Your keys, your plan, the provider's own price. Or pay nothing at all and run free and local models through Ollama.
your keys · your plan · your rate
Your data isn't the product
Nothing collected about how you work, so nothing to sell or train on. The privacy policy is short, go read it.
not harvested · not sold · not trained on
Free for individual developers
Not a trial, and not a seat that disappears when someone rewrites the pricing page. A paid license is only for commercial production or redistribution. Want a bespoke edition? Talk to us.
free forever for individuals · source-available
The whole thing is inspectable, runs locally, and keeps working if we disappear.
| Klaussy | Hosted agent platforms | Single-vendor AI IDEs | |
|---|---|---|---|
| Runs on your machine | Yes | No, their cloud | Yes |
| Uses your own subscription | Yes | Their credits | Their plan |
| Marks up your tokens | Never | Usually | Usually |
| Runs on free or local models | Yes, via Ollama | No | Rarely |
| Works with no AI subscription at all | Yes | No | No |
| Mixes agents in one session | Any of 11 | One at a time | Theirs only |
| Many repos in one session | Yes | Rarely | One window each |
| Reviews the diff before you commit | Built in | No | No |
| Docker-sandboxed agents | Yes | Yes, theirs | No |
| Cost for one developer | Free | Per seat + usage | Per seat |
Bring your own agent. All of them, if you like.
Klaussy drives the agent you choose, it doesn’t replace it. Pick a global default, switch it per tab, or run several side by side in the same session, each on its own worktree. You need at least one installed and authenticated. Klaussy’s first-run check tells you what’s missing.
- Claude CodeAnthropic · the default
- CodexOpenAI
- GeminiGoogle
- AntigravityGoogle
- CopilotGitHub
- CursorAnysphere
- ClineVS Code extension
- opencodeopen source
- Kimi CodeMoonshot AI
- Ollamavia Aider · local models
- Plain shellno agent, just a terminal
Or run any agent in a sandbox
Pick Nemesis8 Sandbox as the agent for any tab and it runs inside an isolated Nemesis8 Docker container instead of directly on your machine, so its blast radius stays off your host. Add a gateway under Preferences → Nemesis8 Sandboxes. One click sets one up locally, or point it at a remote one. Add more than one and each shows up in the agent picker as Nemesis8: <name>, so a sandboxed Claude and a sandboxed Codex can run side by side.
Leave your desk. The session keeps going.
Turn on notifications for a session and it mirrors into Discord or Slack as a thread you can talk to from your phone. Steer the agent while you're out on errands, then walk back to Klaussy and pick the same conversation up mid-sentence. Full scrollback, same worktree, nothing replayed.
Turn on notifications
Switch on notifications for the session and point them at Discord or Slack. Klaussy opens a thread there and streams the agent's output into it as it works.
Drive it from anywhere
Reply in the thread and the agent picks it up on your machine. Answer a prompt, approve a plan, redirect the work, all from your phone in the grocery line.
Come back and continue
Everything said in the thread lands in the Klaussy tab. Sit down and keep typing. One session the whole way through, no handoff brief, nothing lost.
Ran the migration against a scratch DB. 41 tests pass, 2 fail on the legacy token path. Want me to keep the old format readable for one release, or hard-cut it?
keep it readable for one release, put the cutover behind a flag
Done. AUTH_LEGACY_TOKENS defaults on, cutover path tested. Branch is clean and ready for you to review back in Klaussy.
What it does
Run a fleet, not one chat window
One task per worktree, as many agents as you want, all going at the same time.
Parallel agents, isolated worktrees
One task = one git worktree + one agent. The sidebar tracks status, diff counts, and branch drift per task.
Multi-repo sessions, your way
One repo or many, each on the same branch in its own worktree. Clone straight from your recent GitHub repos, resume any session with one click.
Broadcast to all agents
One prompt, every agent in the session, every worktree. Typed once.
Cross-agent session resuming
Pause on Claude Code, resume under Codex. Klaussy writes the handoff brief that catches the new agent up.
Keep them aligned with your code
Every run grounded in your conventions, and nothing lands without a review.
Repo-aware agents
Klaussy maps your conventions, rules, and import graph at session start. Every run works from that instead of guessing, and it refreshes as you go.
Commit-time review gate
Pre-commit and pre-push hooks run an agent over your staged diffs, hunting silent failures, leaked secrets, debug leftovers, and correctness bugs, then tidy what they can.
Plan · Debug · Review
A dropdown on every worktree spawns a dedicated agent tab: a guided Plan flow, a Debug pass, or a multi-phase PR review.
PR reviews without local checkout
Paste a pull request URL for files, commits, checks, and an AI review. Implement, ignore, or append each finding, or pull the review into a worktree.
Auto-debug CI failures
One click on a red check pulls the job logs. Your agent says what broke and drops a suggested fix in the worktree.
On your terms
Sandboxed, local, or driven from your phone. However you want to run it.
Agent mirroring to Discord & Slack
Mirror a session into a Discord or Slack thread and steer it from your phone. See how it works.
Inline AI, locally
Tab-autocomplete powered by qwen2.5-coder via Ollama. ~100ms latency. No code leaves your laptop.
Multi-agent MCP manager
One catalog of MCP servers: GitHub, Slack, Linear, Notion, Datadog. Add one, it fans out to every agent you pick.
Workflow-oriented tools
Fuzzy command palette (Cmd+K), pop-out task windows, and built-in flows for planning, debugging, and humanizing prose.
Built-in editor
Monaco editor with LSP diagnostics. Edit and commit straight from the diff panel. AI-generated commit messages optional.
Enterprise & Partnerships
Bespoke AI development environments. Built for you.
Need a multi-agent coding environment built around how your company already works? Our developers and AI architects work with you directly to fit Klaussy to your workflows, your security rules, and your internal APIs.
Tailored Environments
Extend Klaussy with integrations for your private repositories, internal tools, and proprietary developer systems.
Custom AI Orchestration
Work with our AI architects to set up multi-agent behaviors, model bindings, and security rules for your team.
VPC & Secure Deployment
Host it entirely on your own hardware or inside your private cloud, with strict data governance, custom LLM routing, and the compliance story your auditors want.
FAQ
Does my code get sent to third parties?
When you use an agent, your prompts and repo context go to whichever provider you picked, through the CLI you already trust. GitHub operations go through your local gh. Inline autocomplete runs on your machine via Ollama and qwen2.5-coder:1.5b, so nothing per-keystroke leaves it. We don't run a server of our own.
Do I need a subscription, and does Klaussy mark up AI usage?
You need whatever plan each agent CLI is set up with, on your own accounts. Klaussy never sits between you and your provider, so there's nothing to resell, mark up, or meter. Or run free and local models through Ollama and pay for nothing at all.
What is klaussy-agents?
klaussy-agents is the engine underneath the desktop app: the conventions, custom skills, and git-level guardrails that keep agents aligned with your codebase. The desktop is the face, the engine is the spine. It's open source under the MIT license.
Which platforms are supported?
macOS 12+ (Apple Silicon and Intel), Windows 10/11, and Ubuntu 22.04+ (other modern Linux distros generally work). All builds are signed where the platform supports it.
Why the 2 GB download prompt for inline autocomplete?
~500 MB is Ollama's runtime, ~1 GB is the qwen2.5-coder:1.5b model weights. You only see the prompt if you opt in. Otherwise a free word-based completer handles Tab.
What happens to my data if I uninstall?
On macOS, remove ~/Library/Application Support/Klaussy and ~/Library/Logs/Klaussy. On Windows, remove %APPDATA%\Klaussy. On Linux, remove ~/.config/Klaussy and ~/.local/share/Klaussy. Ollama and its models stick around on their own, so uninstall it through your package manager and remove ~/.ollama.
Is this open source, and what is SUL 1.0?
Klaussy Desktop is source-available under the Sustainable Use License. You can inspect, modify, run, and self-host it for personal use, development, and internal business use. Commercial production use needs a separate agreement, and you can't use it to host a competing service, so get in touch if you want to partner. The engine underneath, klaussy-agents, is fully open source under MIT. Bundled components are listed in About → Licenses.
How do I report a bug or request a feature?
Open an issue on GitHub. Every one gets read.
Get Klaussy.
macOS, Windows, Linux. Source-available under SUL 1.0, free for personal use and development.
Download Klaussy macOS · Windows · Linux