From a school’s question to a donated Mac, and back.
Tokens for Good connects two sides that never meet: a school or a cause that needs AI it cannot pay for, and someone whose Mac has power to spare. Here is the whole route, from both sides.
Five steps, from asking to a beating heart.
To a school it works like the AI tools it already uses. With an API key, the answer is written on a computer that someone has lent to a good cause. Through Claude or ChatGPT, heavy text work goes there, so the school’s own plan lasts longer.
Step 1
A school applies
A school or a cause tells us who it is and how it plans to use AI. We review every application. The first causes are few, and close to home.
Step 2
It asks in the tools it knows
Once accepted, it gets a free monthly budget. With an API key, a donated computer writes the answer in any tool that takes one. Through Claude or ChatGPT, heavy text work such as translating and summarising goes to us, so its own plan lasts longer.
Step 3
A donated Mac answers
The question goes to a donated computer that is giving right now: plugged in and awake. An open model runs there, on the computer itself.
Step 4
The answer goes back
The answer comes back to the tool the question came from. Students never see whose computer wrote it.
Step 5
The heart beats
On the donor’s Mac a small heart in the notch beats while it answers. The app shows that it is helping, never the question. The donor’s dashboard shows which causes.
The same answer, seen from each side.
For a school or a cause
- Apply. Tell us who you are, what your school or cause does and how you plan to use AI. We review each application. The first causes are few, close to home and chosen with care.
- Get access. Once you are in, you get a free monthly budget. Connect it to Claude or ChatGPT, or use an API key with any tool that works with the OpenAI API.
- Ask. Your students and your team ask in the tool they already use: questions, writing help, translating, summarising, sorting texts and pulling facts out of them, on open models.
- Answers come back. Each answer is written by a donated computer. Your students never see whose, and the app never shows the donor the question. Keep sensitive information for the trusted tier.
For someone with a Mac
- Get the app. A small menu bar app for your Mac, with a heart in the notch. Windows and Linux computers can give too.
- Plug it in. It gives whenever your Mac is plugged in and awake. When your Mac sleeps, it waits. Pause whenever you like.
- Forget about it. It runs an open model locally and answers the questions sent to it, quietly in the background. On battery it pauses.
- Watch the heart. It beats while your Mac gives. Point at it in your notch to see that your Mac is helping, and to pause or resume. Your dashboard shows which causes it has helped.
You set the limits. The heart shows it.
There are no pop-ups. It gives quietly in the background, also while you work, and these are the controls you have.
Plugged in
It gives whenever your Mac is plugged in and awake. When your Mac sleeps or the lid is closed, it gives nothing and simply waits.
- Plugged in and awake
- Gives
- Asleep
- Waits
Pauses on battery
By default it only gives while your Mac is plugged in. Unplug it and it pauses. Plug it back in and it picks up again.
- On battery
- Pauses
- Plugged back in
- Gives again
Pause any time
Point at the heart in your notch and press Pause. It takes no new questions until you resume, and the heart stops beating.
- Paused
- No new questions
- To resume
- One click
No notch? Click the heart in your menu bar and choose Pause or Resume. For advanced use, the agent can also be limited to set hours, such as evenings and nights, with an option on the command line.
Press Try Pause and watch the heart at the top of the page stop. Whatever you choose, one principle holds: it should never use more energy than the value it delivers. How that is measured is being decided.
What people ask before they start.
Plain answers. Where something is not decided yet, we say so.
Does it slow my Mac?
It gives in the background, also while you work, and while it answers a question the model uses a fair share of your Mac’s power. During heavy work you may notice that.
If you notice it, point at the heart in your notch and press Pause. It takes no new questions until you resume. For advanced use, the agent can be limited to set hours from the command line.
Does it use my battery?
Not by default. It pauses as soon as your Mac runs on battery, and picks up again when you plug it back in.
Can the model see my files?
The model runs on your Mac through Ollama, and the app gives it no access to your files or apps. It receives the text of a question and sends back the answer, nothing more.
Can I see what students ask?
The app never shows you the questions or the answers. It shows that your Mac is helping, and your dashboard shows which causes it has helped. Students, in turn, never see whose Mac answered.
Today the text is processed in plain form on the donor’s computer. That is why work that any donor can receive (the open tier) is only for information that is not sensitive. More sensitive work goes only to donors under contract, with a data processing agreement (the trusted tier). Stronger isolation is being built now. How we keep it safe
Which models run on my Mac?
Open models, run locally through Ollama. The app picks one by memory: Google’s Gemma 4 (gemma4:26b) on Apple silicon with 32 GB or more, gemma4:12b with 16 GB or more, and qwen3.5:4b on other Macs. They work with text only: answers, writing, translations and summaries.
Which Macs can give?
Macs with Apple silicon are best, because they do the most work for each watt they use. Windows and Linux computers can give too.
What does it cost a school?
Nothing. Accepted schools and causes get a free monthly budget: no seats to buy and no plan per person.
Does it cost me anything to donate? Being decided
What you give is your Mac’s spare power and the electricity it uses while it gives. Whether the app will have any paid option for donors is being decided.
How is the energy rule measured? Being decided
The principle: it should never use more energy than the value it delivers, and efficient chips come first. How that is measured, and how the rule is applied, is being decided.
Which causes does my Mac help? Being decided
The first causes are a few schools and foundations close to home, each reviewed before they get access. Which causes come next is being decided. Schools and causes can join the waitlist now.