Ask your own AI
about your own day
Flow AI is an AI chat for Mac that runs on your own API key or a local model. Ask what needs you today, which repo used the most Claude this week, or what is blocking a launch, and it answers from your FavTray data through read-only tools. It reads only the areas you tick, and FavTray runs no AI server.
Watch it in action
Flow AI Ask: ask about what's on screen. Ask about what is on screen, answered by your own AI provider or a local model.
More videos →Ask beside the numbers, not instead of them
Ask, on any screen
The Ask button in a screen's header, or ⌘/, opens a drawer beside the real numbers. The screen stays visible and clickable while the answer arrives, so you can check it against what is in front of you.
Home chat
Ask about your whole day from Home: what needs you, across GitHub, meetings, projects and AI limits.
The Home brief
A short list of what needs you, worked out on your Mac with no provider at all. Connect one and Flow AI can also turn those rows into a paragraph once a day. It is never counted.
Every answer shows its work
The question, the steps Flow AI took and the tools it read, the answer, anything it couldn't do, then Copy, Regenerate, Edit and the model's name.
Which AI does it use?
Yours. Flow AI talks to any OpenAI-compatible chat completions endpoint. A preset fills in the address and a suggested model, and nothing is locked. Your providers, models and keys are kept together in one macOS Keychain item.
- OpenAI
- DeepSeek
- OpenRouter
- Groq
- Together AI
- Mistral
- Fireworks AI
- Ollama (local)
- LM Studio (local)
- Any OpenAI-compatible endpoint
With Ollama or LM Studio on your Mac, nothing leaves it. With a cloud provider, requests go from your Mac straight to that provider, billed on your own account.
You tick what Flow AI may read
In Settings › Flow AI, each area has a checkbox and an example built from the same fields that would be sent, so the screen can't promise one thing and send another. Before Ask's first send to a provider, one sheet lists what your boxes let through. After that there is no popup on every question.
- Meetings
- Space
- GitHub
- Ports
- System
- AI Usage
- Cloud Costs
- Project Tracker
- Reminders, Eye Rest and Habits
- Always removed: tokens, keys and passwords, including from command lines and environment values, before anything is shown or sent. There is no setting to turn this off.
- Read-only: Flow AI reads through tools that can't change your boards, ports, meetings or files.
- Checkable: every answer lists the tools it read, so you can open the screen and check the number.
What does Free include?
| Free | Pro ($4.99 once) | |
|---|---|---|
| Connect your own provider or a local model | Included | Included |
| The Home brief | Included, never counted | Included |
| Answers in Ask and Home chat | 10 a day, shared; stopped or failed answers don't count | Unlimited |
| Choose which areas Flow AI may read | Included | Included |
The 10 a day resets at local midnight. Past it, earlier answers stay and every screen still shows its numbers. Compare every tool on the pricing page.
Frequently asked questions
What is Flow AI?
Flow AI is the AI chat inside FavTray. It answers questions about your own FavTray data, such as your meetings, pull requests, ports, disk, AI limits and project boards, using your own AI provider and API key or a local model. FavTray runs no AI server and bundles no model.
Which AI providers work with Flow AI?
Any OpenAI-compatible chat completions endpoint. Presets fill in OpenAI, DeepSeek, OpenRouter, Groq, Together AI, Mistral and Fireworks AI, plus Ollama and LM Studio running on your Mac, and Custom endpoint takes any other compatible server. Your key is kept in the macOS Keychain.
Can I use it without sending data to the cloud?
Yes. Point Flow AI at Ollama or LM Studio on your Mac and the questions and the data it reads stay on your Mac. With a cloud provider, the data goes from your Mac straight to that provider, not through FavTray.
What data does Flow AI send?
Only the areas you tick in Settings › Flow AI: Meetings, Space, GitHub, Ports, System, AI Usage, Cloud Costs, Project Tracker, and Reminders, Eye Rest and Habits. The example beside each checkbox is built from the same fields that are sent. Before the first send to a provider, a sheet lists what your boxes let through. Tokens, keys and passwords are always removed first, including from command lines and environment values, and there is no setting to turn that off.
Can Flow AI change my data?
No. Flow AI reads your FavTray data through read-only tools. It can answer questions about your boards, ports or meetings, but it does not edit them.
Is Flow AI free?
Yes. Connecting a provider is free, the Home brief is never counted, and Free gets 10 answers a day, shared by Ask and Home chat. An answer you stop, or one that fails, is handed back. The count resets at local midnight. Pro, a one-time $4.99 Lifetime license, is unlimited, and the 7-day trial is unlimited.
Do I pay for the AI separately?
Yes, if your provider charges for it. Flow AI uses your own API key, so any usage is billed by your provider on your account. A local model in Ollama or LM Studio costs nothing per question.
Rewrite text in any app with the same provider: AI Writer. Track what your AI plans cost and when you hit the limit: AI Usage Tracker. What FavTray stores and where: the privacy policy.
Your data, your model, your answers
Free with 10 asks a day. Pro, $4.99 once, is unlimited. 7-day unlimited trial, no card, no account.