Claude for Nonprofits: The Plugin and Your First Month

Claude for Nonprofits is worth adopting when recurring grant, donor and reporting work already has reliable source material and a named reviewer. Anthropic's program now includes a plugin with 14 skills, alongside discounted access and training. Start with a small, governed rollout, then expand when staff can show that reviewed work gets easier.

What Claude for Nonprofits changes for your organization

Claude, Anthropic's AI assistant, can now begin with nonprofit-specific instructions for familiar operational work. On 26 October 2026, Anthropic released the nonprofit plugin, a package of skills and connectors installed automatically on nonprofit Team and Enterprise plans at no additional cost. A skill is a reusable set of task instructions; a connector gives Claude a way to work with another service. The nonprofit plans use the same models as standard Team and Enterprise and include Claude Code, Anthropic's coding tool, at no additional cost. Anthropic's announcement.

Think of the plugin as an operations handbook for an assistant. It supplies procedures, but your organization still supplies the records, permissions and judgment.

The published inventory covers these areas:

These are the 14 skills listed on Anthropic's nonprofit program page. A polished draft is an intermediate deliverable. Your grant lead still decides whether an opportunity fits the mission; your executive director still owns what goes to the board.

For an executive director, the opportunity is consistent preparation across overloaded roles. Require a clear owner for each output before distributing seats broadly.

For an operations lead, the work begins with access and source ownership. Decide which budget is authoritative, where the current policy lives and who may connect each system.

For a development lead, begin with a recurring grant or donor task whose factual inputs you can already verify. Starting with an unfamiliar prospect and scattered records makes it harder to distinguish research errors from missing organizational knowledge.

The product facts here were checked against Anthropic's public pages for this publication. The workflows below are proposed operating examples, not results from a signed-in trial.

Who qualifies and how to apply

Eligibility is broader than a US charity designation, but it does not include every mission-driven organization. Anthropic lists registered 501(c)(3) organizations and international equivalents, K-12 public and private schools, and these healthcare providers: independent Critical Access Hospitals, Rural Emergency Hospitals, Federally Qualified Health Centers and Rural Health Clinics. Program eligibility.

Its Academy guide excludes government agencies, political campaigns, higher education institutions and healthcare systems. A nonprofit name or social purpose alone is therefore insufficient. Eligibility exclusions.

Anthropic describes the program discount as up to 75% on Team and Enterprise, with approximately 90% for verified nonprofits in low- and middle-income countries; the current Team program card lists Nonprofit standard at $8/user/month and Nonprofit access at $3/user/month for nonprofits based in those countries, excluding applicable tax. Original discount terms, expanded discount terms, current program rates.

Team is advertised for 2-150 seats, with a minimum of 2 seats. The current nonprofit page directs larger organizations to its team and does not state a numerical Enterprise minimum. Ask nonprofit sales to confirm your Enterprise terms rather than importing a minimum from another offer. Nonprofit seat terms.

Apply for the Claude nonprofit discount

For Team, open Anthropic's nonprofit page and select Verify via Goodstack. After approval, use the email instructions to complete setup with the discount applied. For Enterprise, contact nonprofit sales; Anthropic verifies eligibility during that process. Application guidance.

Already paying for an account? Anthropic directs existing customers to Admin settings > Billing or support. Its guide says the discount continues while the organization maintains nonprofit status. Confirm the change on your own account before inviting the rest of your staff. Existing-account guidance.

Connect the records your staff can already use

The nonprofit connectors serve different jobs. Pick the one that reaches the evidence your workflow needs.

• Blackbaud connects Claude to Raiser's Edge NXT, its donor management platform, for donor records, gift history, events and personalized communications. This is the relevant connection when the question depends on your existing donor relationship. Anthropic's connector overview.
• Candid supports nonprofit and foundation discovery, organizational profiles and philanthropic knowledge. Anthropic describes the connector as beta; basic search needs no separate Candid account, while advanced features may require authentication. Its strongest coverage is US organizations. Treat a promising match as a research lead to check against the funder's own requirements. Candid connector tutorial.
• Benevity supports discovery of verified nonprofits by cause, location or keyword. No separate Benevity account is required. Its connector supports research, not donations, so it should not be mistaken for access to your private donor history. Benevity connector tutorial.

Anthropic also names Salesforce on its nonprofit page, and Microsoft 365, Gmail, Google Drive, Asana, Slack and HubSpot in its getting-started guide. Your donor database and your document store may both matter: a gift record cannot explain a program outcome that lives in a separate report. Named integrations.

A data map tells Claude where to look

Begin with /where-things-live, the skill that records where your information is kept, then /org-voice-guide to establish how your organization writes. The map stores locations rather than the contents of your records. Saving it for later chats requires memory and a task running in the cloud; without memory, the map does not persist between chats. Nonprofit setup tutorial.

An owner making a connector available does not automatically connect each person's account. Staff generally authorize the service with their own account, and Claude follows that access. Custom connectors using shared credentials and enterprise-managed authorization need separate scrutiny. Owners can also govern actions, including which require approval. Connector access controls.

If the plugin is missing, have the owner check Organization settings > Plugins & skills and confirm that Skills and Cowork are enabled. Cowork is Claude's mode for working through delegated tasks across files and tools; our Claude Cowork guide explains that workflow. Plugin administration requirements.

Set data handling and admin controls before inviting staff

The relevant question is what happens to a donor record throughout its use, including the conversation containing it.

Training and feedback. Anthropic says Team and Enterprise conversations, files and connected data are not used to train models by default. Its nonprofit tutorial identifies submitted feedback as an exception that may be used to improve Claude. An Owner or Primary Owner controls Rate chats under Organization settings > Data and Privacy. Make that choice explicit in staff instructions. Nonprofit data guidance, feedback setting.

Who can read the work. Chats are private from colleagues unless shared, and owners cannot simply browse everyone's chats in the app. However, the Primary Owner can export organizational conversations. Enterprise's Compliance API can expose conversation and file content to compliance tools. Do not describe work-account chats to staff as inaccessible to the organization. Visibility guidance.

Retention and location. Team conversations remain until deleted; Enterprise supports custom retention, with project retention configured separately. Deleted chats leave the user's history immediately and normally leave Anthropic's back-end storage within 30 days, subject to legal, safety and other documented exceptions. Anthropic's retention page also identifies additional requirements for certain covered models. Retention policy.

Data is stored in the United States, while processing can occur internationally. An organization with a strict location requirement should resolve that requirement before uploading its records. Hosting and processing locations.

Choose Enterprise when the control is the requirement

Team and Enterprise both offer single sign-on, domain verification and just-in-time account creation. Enterprise adds automated account provisioning through SCIM, custom roles, audit logs, the Compliance API and custom retention. These determine who gains access, how departures are handled and what evidence an administrator can retrieve. Anthropic's nonprofit admin guide.

Read our Claude Enterprise guide if those controls determine the purchase. A smaller organization can still have demanding governance needs; staff size alone is a poor substitute for an access requirement.

Before inviting staff, record the approved connectors, sharing rules, retention expectations and the person who handles exceptions. Review project visibility and external artifact sharing as distinct settings. A grant draft approved for an internal colleague may still be inappropriate to share through an external link.

Worked workflows for grants, donors and board reporting

Use Claude to produce something a responsible staff member can check. The examples below describe proposed handoffs, not measured outcomes.

Grant drafting: eligibility before persuasive prose

Situation: A community arts nonprofit is considering a funder's request for proposals, or RFP. Its development lead has the application guidelines, an approved program description, a prior proposal and the current budget.

Start with /funder-and-grant-research to examine fit, then use /grant-proposal-writing for the application draft. These are the skills Anthropic documents for reviewing requirements and drafting from existing material. Fundraising skills.

A useful instruction is:

Compare the application requirements with the documents supplied. Separate confirmed eligibility from missing evidence and disqualifying conditions. Then draft the narrative using only approved program claims. Attach a source to each factual assertion and leave missing information visibly unresolved.

The important output is the requirement list beside the draft. Suppose the proposal describes a wider service area than the approved program document. The reviewer should be able to trace that expansion to a source; persuasive language alone is insufficient.

Human review: The development lead checks eligibility and required attachments against the funder's current application. The program lead approves delivery commitments. Finance confirms the budget. Nobody submits until those owners resolve the gaps. If grant terms introduce legal obligations, use the review boundaries in our Claude for Legal guide and involve the appropriate reviewer.

Donor research: a sourced relationship brief

Situation: A food-access nonprofit is preparing to approach a foundation. The development lead wants to understand the fit without overlooking an existing relationship.

Use approved relationship records alongside public funder material. Candid can support foundation discovery; Blackbaud is relevant where the organization's relationship history lives in Raiser's Edge NXT. Ask for a brief that separates public evidence from your own records and from unresolved questions.

Prepare a relationship brief for this foundation. Explain the connection between its published priorities and our approved program. Distinguish public information from our contact history. Flag conflicting details and anything that needs confirmation. Suggest questions for the conversation, without inventing a personal connection or a funding commitment.

A foundation's published interest in food access is evidence of possible fit. It is not evidence that a program officer knows your organization or that an application is welcome.

Human review: The relationship owner opens the cited pages, checks contact history and confirms outreach is appropriate. Research on named individuals, wealth screening or bulk donor exports should follow the organization's approval rules before data is shared. The final choice of whom to approach remains with the development lead.

Board reporting: explain the variance without inventing the cause

Situation: An operations lead is preparing a board packet using approved financial reports, program updates, the previous meeting's actions and the agenda.

Use the plugin's reporting and deck capabilities together to draft the narrative and presentation outline.

Draft the board update from these approved records. Keep reported outcomes separate from interpretations. Identify the source of every figure and flag differences in reporting periods. Where a result differs from plan, ask the responsible lead for the explanation. Finish with the decisions the board needs to make.

A report may show lower participation while a staff note mentions a venue closure. Claude should not turn those neighboring facts into a confirmed causal explanation. The program lead must establish whether the closure explains the change.

Human review: Finance reconciles figures to the approved accounts; program leads validate outcomes and explanations; the executive director approves the decision requests and confidentiality of the packet. Remove unresolved claims before distribution. Anthropic's tutorial likewise requires source checks and human review for reporting. Reporting review guidance.

Run the first month around reviewed work

Organize the rollout around a deliverable your staff already owe. A real deadline and an existing reviewer make the exercise useful without needing a speculative transformation project.

1. 1

   Begin with ownership and policy

   Name the account owner, workflow owner and output reviewer. Ask staff what they already use AI for and which data they want to use. Decide what is allowed, what needs approval and what stays out. Check the organization's settings against those decisions before sending invitations.

2. 2

   Give the assistant reliable context

   Build and check the record-location map, supply approved examples of your organization's writing and make the signed AI policy available. Enable only the connectors needed for the selected work. A map pointing to last year's draft budget is worse than asking staff for the current file.

3. 3

   Complete a supervised work cycle

   Choose the grant draft, donor brief or board update that best matches your immediate workload. Preserve the inputs, Claude's draft, the corrections and the approved version. Have the usual reviewer explain which errors would have mattered if missed.

4. 4

   Decide whether to expand

   Look at completed, approved work. Did drafting and checking together become easier? Were sources traceable? Did staff keep work within approved systems? Expand only when the workflow owner can answer those questions with examples. Repair recurring source or access problems before adding users.

AI Fluency for Nonprofits

Use the free AI Fluency for Nonprofits course as part of the rollout. Developed with GivingTuesday, it covers describing tasks, assessing outputs, deciding what to delegate and acting responsibly with the result. Staff do not need a technical background. Course and curriculum.

Pair learning with review of an actual draft. Ask staff to explain why an unsupported claim should be removed, or why a private donor note should not appear in a board presentation. That exercise makes the policy concrete.

What's overhyped, and when to wait

The plugin does not make source quality somebody else's problem. Its procedures depend on your records and your review. If staff cannot agree which spreadsheet is authoritative, settle that first.

A connector is not a universal donor-intelligence feed. Candid and Benevity have different research purposes, and an available integration does not establish rights to every record or feature. Match the connection to the question before buying around it.

Faster text is not the adoption result. A grant narrative that takes longer to verify than to write has not helped the development lead. Judge the finished, approved output and the total staff effort needed to reach it.

Do not buy for a task with no accountable reviewer. Also wait if required security controls remain unresolved, if the work is too occasional to justify maintaining a shared setup, or if the hoped-for benefit is autonomous decisions about donors or people receiving services.

Free AI for nonprofits: start with public material

If you are still deciding whether the work fits, start with the free course and a public-facing writing exercise. Anthropic's nonprofit policy tutorial also offers an AI policy template that works with a free Claude account, plus a Google Docs alternative. Policy resources.

That is a useful starting point for a lone operator. Do not purchase unused seats just to qualify for a program. When shared staff work, consistent administration and approved access become necessary, revisit the organizational plan. Keep confidential organizational records out of an exploratory personal-account exercise.

Questions nonprofit leaders ask

Is there a Claude plan for nonprofits?

Yes. Eligible organizations use Team or Enterprise under Anthropic's nonprofit program, with the nonprofit plugin and training resources. Choose based on workflow and administrative requirements.

Does Claude offer a discount for charities?

Yes, for qualifying organizations. Use the Goodstack application for Team or the nonprofit sales route for Enterprise, and confirm the applicable terms during verification.

What is the best AI for non-profits?

Choose around a recurring task, the records it needs and the person who can verify the output. Claude is a strong candidate for document-heavy nonprofit operations when those conditions are met. An occasional public-writing task may need a smaller starting point.

Your Monday move: Bring the operations lead and the relevant reviewer together around an upcoming deliverable. Agree on its sources, permitted data and approval owner, then use that work to decide whether Claude deserves a wider role.

For practical guides to adopting AI at work, subscribe to the newsletter.