Tokens for Good review (2026)

A candid look at Tokens for Good, a way to put your spare Claude capacity to work researching and vetting nonprofits for a public directory. It is a good fit if you already pay for Claude and have headroom you do not use; it is not for everyone, and we say so below.

What it is and how it works

Tokens for Good lets developers contribute idle AI capacity, the model tokens your Claude subscription already includes, to research high-impact nonprofits. The vetted results become a public directory at fiercephilanthropy.org/directory. You install an npm package, connect a remote MCP server inside Claude Code, and optionally set a schedule so research runs on its own. Under the hood it uses a v3 dual-research model: every organization is researched twice by independent contributors, run through deterministic scoring and a validation layer that checks for hallucinated claims, then consolidated into one canonical report. See how research works for the full pipeline.

What is good about it

A few things stand out. First, there is no extra cost; it runs on capacity you are already paying for, so you are not adding a bill or spending on API keys. Second, it is close to set-and-forget: setup takes about a minute, and with the /schedule automation your AI can pick up work without you prompting it each time. Third, the output is verified rather than a single unchecked pass, thanks to the double research, scoring, and anti-hallucination checks. Fourth, and worth stating plainly, this is not crypto; "tokens" here means AI model tokens (your Claude usage), not a coin, wallet, or blockchain. Finally, it is native to Claude Code: npm install plus a remote MCP connection, no separate accounts or provider keys to juggle.

Who it is for, and who it is not

It suits developers who keep a Claude subscription and routinely leave capacity on the table; contributing that headroom costs you nothing you would otherwise use, and the work goes to a cause. It also fits people who like the idea of giving via effort and compute rather than cash. It is a weaker fit if you regularly max out your usage, since research runs would compete with your own work. It is also not for you if you want to give money directly; this is a way to contribute capacity and vetting labor, not a channel for cash giving. And if you are uneasy about your model producing anything that goes public, read the next section first.

What to know before you start

Two practical points. The research draws on your subscription capacity, so on a busy day you may notice scheduled runs using headroom you wanted; the schedule is adjustable, so tune the frequency to fit. And the output is public: your reports feed a directory people read, which is the point, but it means the work is quality-checked and consolidated rather than published verbatim. That review step is a feature, though it does mean your first pass is not the final word. On privacy and safety, the tasks are narrow and structured, and a validation layer exists specifically to catch unsupported claims before they land. If you want to confirm the project is real before committing, we cover that in is Tokens for Good legit.

Verdict and how to start

For developers with spare Claude capacity and an interest in vetting nonprofits, Tokens for Good is an easy recommendation: real cost is near zero, setup is quick, and the verification pipeline means your contribution is credible rather than throwaway. It earns a caveat mainly for people who fully use their capacity or who wanted to give cash; for them it is simply the wrong tool. If it fits, start by reading what it is or jump straight to the setup guide, install the package, connect the MCP server, and pick a schedule.

Frequently asked questions