Claude for Nonprofits: the $8 Seat, the $3 Tier and the Limits

Claude, made by Anthropic, is available to eligible nonprofits at $8 per user per month on its Team plan, against a list price of $20 billed annually or $25 monthly, and at $3 per user per month for nonprofits based in low or middle income countries. The offer, called Claude for Nonprofits, launched on Giving Tuesday, 2 December 2025, and verification runs through Goodstack in about three minutes for organizations under 20 people. This page sets out the prices as Anthropic publishes them today, how the offer compares with ChatGPT’s, what Claude is good for in a nonprofit, the training and privacy settings that matter, and where a general assistant should stop.

What Claude costs a nonprofit

Four details decide whether this goes smoothly.

• The discount is on Team and Enterprise, not personal plans. Pro and Max are not discounted. An organization paying for several individual Pro subscriptions pays list price for an arrangement with no admin control, so two or more people using Claude for work should be on a Team workspace.
• Team needs at least two seats and covers up to 150. Above that, or for Enterprise, the route is Anthropic’s sales team, which verifies status directly.
• Verification depends on size. Anthropic’s page says organizations with fewer than 20 people verify through Goodstack, the same validator used by OpenAI and many other vendors, then receive an email to finish setup at the discounted price. Larger organizations contact sales. How Goodstack decides eligibility is covered in the nonprofit software discounts guide.
• The $3 tier is real and rarely mentioned. Nonprofits based in countries the World Bank classes as low or middle income pay $3 a seat. For an international NGO with a field office, check which entity is applying.

At $8 a seat, a five person team costs $480 a year; at the $3 rate, $180. Anthropic’s announcement described the offer as “up to 75%” off; on the Team prices it publishes today, $8 is 60% below the $20 annual list price and 68% below the $25 monthly one, and $3 is 85% to 88% below. Prices exclude tax, and Anthropic says they may change. Its tutorial adds that a tax exemption certificate can be submitted for a refund of sales tax.

Who is eligible

Anthropic’s nonprofit page lists registered 501(c)(3) organizations and their international equivalents, K-12 public and private schools, and certain mission-based healthcare providers: independent Critical Access Hospitals, Rural Emergency Hospitals, Federally Qualified Health Centers and Rural Health Clinics. Its getting-started tutorial lists the usual exclusions: government agencies, political campaigns, higher education and larger healthcare systems, which have their own programmes. The discount continues for as long as the organization keeps its nonprofit status.

Fiscally sponsored projects are the common edge case. Goodstack validates the legal entity, so a sponsored project usually applies under its sponsor’s registration or not at all. Ask the sponsor before applying.

Claude or ChatGPT for a nonprofit

The two offers are close to identical on price, and both use Goodstack. The differences are elsewhere.

The connectors are the most nonprofit-specific part of Anthropic’s offer. Anthropic launched Claude for Nonprofits with connectors to Benevity (a database of more than 2.4 million validated nonprofits), Blackbaud’s Raiser’s Edge NXT and Candid (data on more than 1.9 million nonprofits and foundations, including grants), alongside general ones for Microsoft 365, Google Drive, Gmail, Slack, Asana and HubSpot. For grant research in particular, asking Claude to query Candid is a real time saver, provided someone checks the funder’s own guidelines before writing a word.

Beyond that, choose on what your team already uses. If staff are already comfortable in one, the second $8 seat is rarely worth it. Our ChatGPT for nonprofits guide covers the other side in the same structure.

What Claude is good for in a nonprofit

• Long documents. Claude handles long inputs well: a 60 page evaluation into a two page board summary, a funder’s guidelines into a checklist, three years of annual reports into a consistent narrative.
• First drafts with your structure. Grant sections, appeal letters, thank-you letters, job descriptions and policies, written into headings you supply. Give it the template and three paragraphs of your own writing for voice.
• Files you can use. Team includes file creation, so a budget narrative can come back as a document and a data table as a spreadsheet.
• Thinking with structure. The objections a board will raise, the questions a funder might ask, a first risk register.

The cautions are the same as for any language model. The output is fluent whether or not it is right, so anything factual must be checked against a source you control, and the default voice is bland unless you give it your own examples.

The training and privacy settings that matter

On Free, Pro and Max, you choose whether your chats train Anthropic’s models. Anthropic’s consumer terms update of 28 August 2025 asked every consumer user to decide; the setting is called Model Improvement, under Privacy Settings. If you allow training, data is kept for up to five years; if you do not, for 30 days. Anthropic’s pricing page describes model training on consumer plans as opt-out, so check the setting rather than assume. Incognito chats are never used for training.

On Team and Enterprise, Anthropic does not train on your data by default. Those plans are outside the consumer terms. Team adds single sign-on and central billing; Enterprise adds SCIM, audit logs, a Compliance API, custom data retention and a HIPAA-ready configuration with a business associate agreement. Anthropic lists SOC 2 Type I and Type II and ISO 27001 certification for its commercial products.

So the practical rule is the same as for ChatGPT: buy the workspace, not individual subscriptions, and the training question is settled. What is not settled by the workspace is the next section.

Where a general assistant should stop: your donor file

Drafting text is one job. Asking a general assistant what is in your donor database is a different job, and three things remain true on a paid Team workspace.

• Figures are assembled in conversation. With a CRM connector, Claude can pull donor records, but totals, rankings and comparisons are put together by the model inside the chat. They may be right; they are not the output of an auditable query, and a wrong lifetime giving figure in a board paper costs credibility.
• Everything you paste or retrieve goes in whole. Names, addresses, gift amounts and the note about a donor’s illness. Not training on it is not the same as minimising it first. Our AI data security guide sets out the questions to ask.
• Nothing happens until someone asks. A chat assistant answers the question you type. It does not review the file overnight and tell you that a loyal donor’s pledge has lapsed.

The third and fourth rows are where Gratefully, an AI donor intelligence and stewardship platform built for nonprofits, sits, and it also works inside Claude. It reads Salesforce for Nonprofits, Raiser’s Edge NXT, Bloomerang, Little Green Light or HubSpot, plus documents and email engagement, into a knowledge graph; every answer from its assistant, Grace, cites the record it came from; personal details are redacted before anything reaches a model; and it works through the whole portfolio each night to produce a ranked list for the morning, drafting outreach that a person approves and sends from their own mailbox. Added to Claude as a connector, it lets you ask Claude “who should I call this week, and why?” or “prep me for coffee with a donor tomorrow” and get answers cited back to Gratefully’s records, and Claude can log a note, set a reminder or save a draft for review, but never sends email or changes your CRM through it. A Claude plugin adds ready-made meeting prep, lapsed-donor win-back and year-end appeal workflows. Gratefully is free for one person and $79 a month billed annually for Essential, and connector usage counts toward the organization’s plan. It is wrong for two cases. If your need is drafting, the $8 Claude or ChatGPT seat does that job and a specialist tool is the wrong purchase. And if your board requires an issued SOC 2 Type II report before donor data is connected, Anthropic has one and Gratefully expects its own in the fourth quarter of 2026.

Many organizations end up with both: a general assistant for writing, and a donor tool for the file. The AI fundraising tools roundup compares the specialist options.

AI Fluency for Nonprofits, the free course

Anthropic built a free course with GivingTuesday, AI Fluency for Nonprofits, on its “4D” framework: delegation, description, discernment and diligence. Coursera lists it at five modules and about six hours, beginner level. The examples use Claude, but the method applies to any assistant, which makes it a reasonable first course for a whole team before buying seats. Other free and paid options are compared in AI training for nonprofits.

Setting it up without causing a problem later

1. Verify, then buy Team. Two seats minimum. Move anyone on a personal Pro plan into the workspace.
2. Write the one page policy first. What may and may not be pasted, who checks facts, and who approves external use. The AI use policy template is the short version.
3. Name the forbidden categories. Identifiable donor records, beneficiary case notes, anything about a child, health information, HR files.
4. Connect deliberately. Each connector widens what Claude can read. Turn on the ones you need, with the narrowest permissions, and record them in the policy’s tool register.
5. Train before rollout. An hour on the free course does more than any prompt list.

How this page was put together

Prices, eligibility and verification were read on Anthropic’s nonprofit page and pricing page on 9 October 2026. Anthropic’s own getting-started tutorial gives an Enterprise figure of $10 a seat that its current nonprofit page does not; we report the current page. The launch date, partners and connectors are from Anthropic’s announcement of 2 December 2025; training rules from its consumer terms update and privacy centre; course details from Anthropic’s course page and Coursera. ChatGPT figures are from our ChatGPT guide, checked the same week. Gratefully is a company we cover closely; its details are from its own site.

This is reference information, not legal or tax advice. Rules vary by state and change over time. For a decision that carries real consequences, check the current text at irs.gov or your state registry, and talk to a nonprofit attorney or CPA.