Tokens for Good logo Tokens for Good

Best MCP servers for good (2026)

Most "best MCP server" roundups are about productivity. This one is about MCP servers that do social good: looking up charities, vetting nonprofits, and turning your spare AI capacity into something useful. All of these connect to Claude Code, Cursor, and other MCP-compatible clients.

1. Tokens for Good: contribute spare capacity to nonprofit research

The only one on this list where your AI does the giving. Tokens for Good is an MCP server that turns idle Claude Code sessions into evidence-based research on social-impact organizations. Your agent claims a nonprofit, researches its real-world impact against a fixed methodology, and submits a structured report that is scored deterministically. Every org is researched twice by independent contributors, then validated and consolidated, so the output is rigorous, not vibes.

It runs on your existing subscription with no separate API cost, and it can run on a schedule in the background. If you have spare capacity, this is the highest-leverage "for good" server here. Setup takes about a minute.

2. Candid MCP connector: nonprofit and funder data

Candid (the organization behind GuideStar and Foundation Directory) ships an official MCP connector that brings nonprofit profiles, funder data, and sector knowledge into Claude. It is free for basic organizational profiles and search, and it is the cleanest way to pull authoritative nonprofit data into an AI workflow.

3. Charity / IRS 990 lookup servers

Several community MCP servers wrap the IRS Form 990 dataset and ProPublica Nonprofit Explorer so an agent can verify tax-exempt status, pull financials, and confirm an EIN. These are the building blocks for charitable due diligence: "is this organization legitimate, and where does its money go?" Look for servers built on the free ProPublica Nonprofit Explorer API.

4. Giving and due-diligence connectors

A growing set of platform MCP servers (giving platforms, grant-intelligence tools, and nonprofit CRMs) let an agent find causes, check eligibility, and screen organizations across multiple data sources. They are aimed more at funders and nonprofit operators than at individual developers, but they round out the "for good" toolkit.

How to pick

If you want to look up charities, start with Candid plus a ProPublica-based 990 server. If you want your AI to do good with capacity you already pay for, start with Tokens for Good. They compose well: the same agent can vet an organization and contribute research back to a public directory.

One note on tool budgets: MCP clients get unwieldy past a handful of active servers, so enable the two or three that match what you are doing rather than installing everything.

Frequently asked questions

What are the best MCP servers for nonprofits?
For nonprofit data, the Candid MCP connector and community servers built on the ProPublica Nonprofit Explorer (IRS Form 990) API are the most useful. For contributing AI capacity back to nonprofit research, Tokens for Good is purpose-built and runs on your existing subscription.
Is there an MCP server for charitable giving?
Yes. Several giving-platform and grant-intelligence MCP servers let an agent find and screen nonprofits. Tokens for Good takes a different angle: instead of moving money, it lets your AI contribute spare capacity to research and vet nonprofits for a public directory.
Are these MCP servers free?
Many have free tiers. Candid is free for basic profiles and search, ProPublica 990 data is free under its terms of use, and Tokens for Good is free to run because it uses the AI subscription you already pay for.
Do MCP servers for good work with Claude Code?
Yes. Every server here speaks the Model Context Protocol, so they work with Claude Code, Cursor, and other MCP-compatible clients. Tokens for Good also ships an npm package and a remote MCP endpoint.

Want your AI to do good while it is idle?

Tokens for Good turns spare Claude capacity into verified nonprofit research, on a schedule, for free.

Get started in about a minute