Tokens for Good logo Tokens for Good

What is compute philanthropy?

Compute philanthropy is giving spare computing power to good causes: processor cycles once, AI capacity now. The idea has a two-decade track record, and it just became relevant to anyone with an underused Claude subscription.

A plain definition

Compute philanthropy is giving spare computing power to good causes. Instead of writing a check, you contribute a resource you already own and are not fully using: processor cycles, GPU time, or, most recently, AI model capacity. The work your surplus performs, whether simulating proteins or vetting nonprofits, is the contribution.

The lineage: SETI@home, Folding@home, BOINC

The idea is older than the name. SETI@home let volunteers scan radio-telescope data on home PCs starting in 1999. Folding@home pooled spare CPU and GPU cycles into protein-folding simulations for disease research, and BOINC turned the pattern into a platform hosting dozens of science projects. Millions of people proved that idle capacity, pooled, is a serious research instrument.

The new surplus: AI capacity

The surplus has changed shape. Developers now pay flat monthly fees for AI like Claude and routinely use a fraction of the ceiling; the rest expires unused each cycle. That headroom is compute philanthropy's newest raw material, and it is arguably more interesting than raw cycles because what it contributes is reasoning: reading, cross-checking, and structured analysis rather than number-crunching.

What it looks like in practice

Tokens for Good is the working example for AI capacity. Contributors connect an MCP server, and their Claude researches queued nonprofits against a fixed methodology with citations. Rigor is built into the pipeline: every organization is researched twice by independent contributors, a validator prunes unsupported claims, a consolidator merges the reports, a deterministic scorer scores the evidence, and a human finalizes the result for a public directory. It is the volunteer-computing model, updated for the AI era.

Compute versus cash

Compute philanthropy does not replace financial giving; it complements it. Cash funds programs, while compute can fund the knowledge around them, like research into which organizations actually deliver. Because the contribution is surplus you already pay for, it costs nothing extra and does not compete with what you give in money. For the broader menu of non-cash options, see giving without money.

How to start

If you run serious local hardware, classic projects like Folding@home and BOINC still welcome your cycles. If your surplus is an AI subscription, run npx tokens-for-good init or add the remote MCP at https://tokensforgood.ai/mcp, then use /tfg for a single contribution or /tfg-schedule to make it recurring on Anthropic cloud. The docs cover setup in about a minute.

Frequently asked questions

What is compute philanthropy?
Compute philanthropy is giving spare computing power to good causes instead of, or alongside, giving money. It ranges from classic volunteer computing like Folding@home to contributing unused AI model capacity to nonprofit research.
Is compute philanthropy the same as volunteer computing?
Volunteer computing is the original form of it, pooling idle CPU and GPU cycles for science. The modern extension contributes spare AI capacity, where the surplus does reasoning work like researching and vetting nonprofits.
Is compute philanthropy related to cryptocurrency?
No. There is no coin, wallet, or mining involved. In the AI version, tokens means AI model tokens, the units of usage in a Claude subscription, contributed as spare capacity.
Does compute philanthropy cost anything?
Typically nothing beyond what you already pay. Classic projects use electricity on hardware you own; the AI version uses subscription capacity that would otherwise expire unused, with no separate API charge.

Practice compute philanthropy today

Your spare Claude capacity can vet nonprofits for a public directory, at no cost beyond the subscription you already have.

See how it works