Plan — Local — 一尾
Free
One tail, held close. Everything stays on your device.
No account required.
Kyubi — Screen-aware assistant
Master anything instantly
Help with anything, anywhere on your computer.
01 — In place
Answers on top of your work
Ask about a menu, a window or a document. Kyubi replies where you are looking.
02 — Private
Your screen stays yours
Screen reading runs on your device. Captures are never stored on our servers.
Read the privacy policy03 — When you need more
Cloud models, on your terms
Start local and free. Add deeper reasoning when you want it, and pay for what you use.
九尾 — Kyubi
Named for the nine-tailed fox
In Japanese folklore a fox grows wiser with every tail. At nine, it sees everything.
04 — Plans — 尾
A tail for every need
In folklore a fox grows a tail as its power grows. Choose how far Kyubi reaches.
Compare every featurePlan — Cloud — 三尾
€7 / month
Three tails, reaching into the cloud when you ask.
Billed annually, or €9 month to month. $5 of cloud credits a month.
Plan — Pro — 九尾 — Most popular
€17 / month
All nine tails. Every model, full reasoning.
Billed annually, or €22 month to month. $20 of cloud credits a month.
Download — Windows
Sharpen your thinking
Free to start, flexible as you grow. Version 0.0.6, early access for Windows 10 and 11.
Pricing — Plans
Your screen, understood
Screen reading runs on-device and never leaves your machine. Add cloud models when you need deeper reasoning.
Plan — Local — 一尾
Free
One tail, held close. Everything stays on your device.
No account required.
- Screen reading & analysis (on-device only)
- Unlimited local model inference
- Ollama & llama.cpp integration
- Local vision models (LLaVA, Moondream)
- Conversation history (local)
- Model manager (pull / delete)
- Screen data never leaves your machine
- Cloud model access (not included)
- Included cloud token credits (not included)
- Send screen context to cloud models (not included)
- Reasoning & planning models (not included)
Plan — Cloud — 三尾
€7 / month
Three tails, reaching into the cloud when you ask.
Billed €84 a year, saving €24. $5 of cloud credits a month.
- Everything in Local
- $5 cloud token credits / mo included
- Send screen context to cloud models (opt-in per session)
- Standard text models (GPT-4o mini, Gemini Flash, Claude Haiku)
- Cloud vision models (GPT-4o, Gemini Flash)
- Conversation sync across devices
- Pay-as-you-go top-ups
- Reasoning & planning models (not included)
- Code-specialist models (not included)
- API key & remote access (not included)
- Priority routing (not included)
Plan — Pro — 九尾 — Most popular
€17 / month
All nine tails. Every model, full reasoning.
Billed €204 a year, saving €60. $20 of cloud credits a month.
- Everything in Cloud
- $20 cloud token credits / mo included
- Reasoning models (o3, DeepSeek R1, QwQ) for deep analysis
- Code-specialist models (DeepSeek Coder, Codestral)
- Frontier vision models (GPT-4o, Claude 3.5 Sonnet, Gemini 2.0)
- Priority routing (lower latency)
- API key & remote access
- Bring your own provider keys
- Model fine-tuning access
- Early access to new models
- 3 concurrent model slots
Screen data is never stored on Kyubi servers, on any plan. See top-up packs
01 — Top-ups
Top up any time
Credits work across every cloud provider and never expire. Larger packs include bonus credits.
$550 credits
$15160 credits — +30 bonus
$30300 credits — +90 bonus
$60740 credits — +180 bonus
$1202300 credits — +600 bonus
02 — Models
What you can run
Local models run on your hardware at no cost. Cloud models draw on your credits.
Browse the full catalogueText & Chat — 9 models — Local, Cloud, Pro
| Model | Provider | Runs | Plans |
|---|---|---|---|
| Llama 3.3 70B | Meta | Local | Local, Cloud, Pro |
| Mistral 7B / 22B | Mistral | Local | Local, Cloud, Pro |
| Phi-4 | Microsoft | Local | Local, Cloud, Pro |
| Gemma 3 | Local | Local, Cloud, Pro | |
| GPT-4o mini | OpenAI | Cloud | Cloud, Pro |
| Claude Haiku 3.5 | Anthropic | Cloud | Cloud, Pro |
| Gemini 2.0 Flash | Cloud | Cloud, Pro | |
| GPT-4o | OpenAI | Cloud | Cloud, Pro |
| Claude Sonnet 4 | Anthropic | Cloud | Cloud, Pro |
Vision & Screen Reading — 6 models — Local, Cloud, Pro
| Model | Provider | Runs | Plans |
|---|---|---|---|
| LLaVA 1.6 | LLaVA | Local | Local, Cloud, Pro |
| Moondream 2 | Vikhyat | Local | Local, Cloud, Pro |
| GPT-4o (vision) | OpenAI | Cloud | Cloud, Pro |
| Gemini 2.0 Flash (vision) | Cloud | Cloud, Pro | |
| Claude 3.5 Sonnet | Anthropic | Cloud | Pro |
| Gemini 2.0 Pro (vision) | Cloud | Pro |
Reasoning & Planning — 5 models — Pro
| Model | Provider | Runs | Plans |
|---|---|---|---|
| DeepSeek R1 (local) | DeepSeek | Local | Pro |
| QwQ-32B | Qwen | Local | Pro |
| o3 mini | OpenAI | Cloud | Pro |
| DeepSeek R1 (full) | DeepSeek | Cloud | Pro |
| Gemini 2.0 Flash Thinking | Cloud | Pro |
Code Specialist — 4 models — Pro
| Model | Provider | Runs | Plans |
|---|---|---|---|
| DeepSeek Coder V2 | DeepSeek | Local | Pro |
| Qwen2.5 Coder 32B | Qwen | Local | Pro |
| Codestral 25.01 | Mistral | Cloud | Pro |
| GPT-4o (code mode) | OpenAI | Cloud | Pro |
03 — Compare
Everything, side by side
| Feature | Local | Cloud | Pro |
|---|---|---|---|
| Screen Reading | |||
| Screen capture & analysis | On-device only | On-device only | On-device only |
| Send screen to cloud models | — | Opt-in per session | Opt-in per session |
| Screen data stored on servers | — | — | — |
| Active window detection | ● | ● | ● |
| Inference | |||
| Local model inference | Unlimited | Unlimited | Unlimited |
| Ollama / llama.cpp support | ● | ● | ● |
| Concurrent model slots | 1 | 1 | 3 |
| Cloud | |||
| Cloud model access | — | ● | ● |
| Included credits / mo | None | $5 | $20 |
| Pay-as-you-go top-ups | — | ● | ● |
| Priority routing | — | — | ● |
| Models | |||
| Text & chat | Local only | Local + Standard | Local + All cloud |
| Vision & screen reading | Local only | Local + Standard | Local + All cloud |
| Reasoning / planning | — | — | ● |
| Code specialist | — | — | ● |
| Platform | |||
| Conversation sync | — | ● | ● |
| API key & remote access | — | — | ● |
| Bring your own provider keys | — | — | ● |
| Model fine-tuning access | — | — | ● |
| Early access to new models | — | — | ● |
| Support | |||
| Support channel | Community | Priority email | |
04 — Enterprise
Enterprise, on your terms
Deploy Kyubi on-premise so screen captures never leave your network. Set org-wide policies for which applications Kyubi can observe, with a full audit log. Contracts include SSO, SCIM and volume token pricing on a single invoice.
On-premDeployment option
CustomToken volume pricing
Air-gappedOffline install available
4 hCritical incident SLA
On-premise deployment
Run the full Kyubi stack on your own infrastructure. Air-gapped installs are available for regulated environments where screen data cannot touch external networks.
Dedicated cloud routing
Your cloud API requests route through isolated infrastructure with a guaranteed throughput SLA. No shared tenancy, no data co-mingling.
Screen data governance
Define org-wide policies for which apps and windows Kyubi may capture, with a full audit log of every screen-context request.
Centralised API keys
One set of provider API keys for your whole org, with per-team and per-user spend limits.
SSO & SCIM
SAML 2.0 / OIDC single sign-on and automated user provisioning via SCIM. Works with Okta, Entra and Google Workspace.
Volume token pricing
Commit to token volumes at discounted rates, unified across all cloud providers on a single invoice.
05 — Questions
Common questions
Does Kyubi ever store what's on my screen?
Never. With a local model, screen captures are processed on your device and never saved to Kyubi's servers. With a cloud model, the capture goes straight to the model and is then deleted; Kyubi does not keep a copy.
How does screen reading work on Local?
Kyubi captures the active window or a selected region, sends it to a local vision model (LLaVA 1.6 or Moondream 2) running entirely on your machine, and the model replies. No internet connection is required and no data ever leaves your device.
What are token credits and how do they work?
Cloud model providers (OpenAI, Anthropic, Google, etc.) charge per token, roughly per word processed. When you send a message or a screen capture to a cloud model in Kyubi, the cost is deducted from your credit balance. Included credits refresh each billing period. You can top up manually at any time and those credits never expire.
Do unused monthly credits roll over?
Included monthly credits do not roll over; they reset each billing period. Manually purchased top-up credits are permanent and never expire. Kyubi always draws from your included credits first before touching your top-up balance.
Which local models can I run?
Any model compatible with Ollama or llama.cpp. Hardware is the main constraint: larger models need more VRAM or Apple unified memory, so check compatibility first.
Can I bring my own API keys for cloud providers?
No, we use proprietary infrastructure and do not support bringing your own API keys for cloud providers.
What happens when I run out of cloud credits?
Cloud model requests will be paused until you top up or your next billing period begins. Local models and on-device screen reading are completely unaffected and will keep running regardless of your credit balance.
Can I switch plans at any time?
Yes. Upgrades take effect immediately and are prorated. Downgrades apply at the end of the current billing period so you keep your current features until then.
06 — Start
Start reading your screen today
Free and on-device. No account, no cloud required.
Blog — Writing
Writing from the team
Software, design, engineering, and building a product people want to use.
Loading posts…
Blog — Letters
Stay in the loop
New posts, product updates and the occasional deep dive. Unsubscribe any time.
About — Kyubi
Tools for people who think
Most AI is built to do the work for you. Kyubi works with you, then steps aside.
01 — Why
Learn anything instantly
AI has shown huge promise, and it is often misused. While companies race to put it in everything, we asked a different question: how can AI make people better at their own work?
Kyubi is our answer. Ask it for help with menus, unfamiliar software, or a second look at your work, and it answers in place, on your screen, in the workflow you already have. The idea is simple: amplify people with AI, do not replace them.
02 — Values
What we believe
Practicality
We look for practical uses for AI. If we use a technology, it is the best fit for the job.
Built for compliance and security
Kyubi is built from the ground up with industry standards in mind and is compliant with all major data protection regulations.
Innovation and AI democratisation
We believe in pushing AI forward, and in making that power accessible.
Built to last
We make software that is built for the future.
03 — People
The team
Co-founder & CEO
Michal
Studying theoretical and applied mathematics and computer science, passionate about machine learning.
Co-founder & Head of Design
Kai
Previously a law student, and has worked on and off in the VFX industry for three years.
Joining soon
Dhruva
Waiting on a visa.
Support — Help center
How can we help?
Search the common questions, or reach out directly.
Questions
Common questions
What is Kyubi?
Kyubi is a free and flexible desktop application designed to sharpen your thinking and accelerate your abilities. It provides tools and workflows to help you learn, create, and perform at your best.
Which platforms does Kyubi support?
Kyubi currently supports Windows (x64). macOS and Linux builds are in active development; see the roadmap to track progress.
Is Kyubi really free?
Yes. Core features are free and always will be. Some advanced features come with optional paid plans.
How do I install Kyubi on Windows?
Download the installer from download.kyubi.dev, run the .exe file, follow the setup wizard, and launch Kyubi from your Start menu or desktop shortcut.
Still need help?
We reply within one to two business days.
DiscordFastest responseEmailsupport@kyubi.devOpen the contact page