Grantable

Free has no time limit and needs no card. The pricing page includes funder search, AI chat, documents and sharing, but caps AI chat at 5 messages a day and file imports at 10 MB. The limit resets daily, so Free suits a short daily check of funders or a single section draft rather than a full application.

Each message in a long chat replays the earlier conversation to the model, so a long thread uses more budget than several short ones. Upload the proposal or report you want to draw on before you start, and ask for one specific output per message.

Paid plans are described relative to Free: Starter roughly 20 times its AI capacity and Pro roughly 60 times. The pricing FAQ calls these approximate guidelines, not hard limits. The only absolute figures are chat messages, 50 a day on Starter and 200 on Pro, and unlimited file imports on both.

Behind the messages sits a monthly AI budget metered by model tier and conversation length: Fast uses Claude Haiku and costs least, Auto and Pro use Claude Sonnet, and Pro costs most per message. When the budget runs out, AI features pause while files stay available; you can buy an overage pack or upgrade.

Qualifying nonprofits get 50% off Starter or Pro for 12 months: $25 or $75 a month. Eligibility is a 501(c)(3) with an annual operating budget under $500,000, confirmed by emailing the latest IRS Form 990. The discount can be renewed with a new 990, and on an existing paid plan it starts at the next billing cycle.

Editorial assessment

For a small nonprofit writing a few applications a quarter, the half-price Starter year is the clear step up from 5 messages a day. Pro makes sense when several grants run at once or scheduled scans matter. The pages do not say whether the discount stacks with annual billing, so ask before paying a year upfront.

The pricing page says there are no feature gates and that every plan includes agents, writing, dashboard, scans and skills. The help page on plans, updated July 30, 2026, draws a different line: scheduled tasks, deep funder intelligence and the Pro model tier are listed only on Pro and a Pro+ tier, and extra usage cannot be bought on Free.

Pro+ does not appear on the current pricing page. Until the two pages match, treat scheduled scans and the higher-quality model tier as Pro features, and confirm access in the app before choosing Starter for automated monitoring.

The terms say you own your content and AI output and that your data is not used to train AI models. The sub-processor list, updated September 27, 2026, names Anthropic as the only AI provider that reads what you write or upload, with workspace data stored by Supabase on AWS in North Virginia. The pricing FAQ and terms name further model providers, so the lists do not fully agree.

After you ask to delete an account, the terms allow 30 days to export content in a machine-readable format and then deletion from production systems within 90 days. Cancelling a plan is not deletion: workspaces and documents stay on the Free plan until you remove them in account settings.

The terms describe AI output as assistive, not authoritative: it may contain inaccuracies, and you must review and verify it before submitting, particularly in grant applications. They also state that Grantable does not guarantee funding outcomes and that the accuracy of every submitted application is your responsibility.

Some funders set their own rules on AI-assisted applications, so check the guidelines before drafting. Keep each statistic tied to a source document in the library, and compare the RFP checklist with the funder text rather than trusting the checklist alone.

1. Upload the source document and paste the RFP requirements.
2. Message 1: ask for the RFP checklist.
3. Message 2: ask for a 300-word needs statement using only the uploaded document.
4. Messages 3 and 4: request one revision and a list of every figure with its source page.
5. Check each figure against the document and each checklist item against the RFP.

Accept a statement whose figures all trace to the source and a checklist that matches the RFP. Reject invented statistics, missed requirements or a draft that needs more than the daily limit. Leave the observed result blank until the exercise is run.

For general drafting without funder research, compare Claude and ChatGPT. For checking statistics against published studies before they go into a proposal, see Elicit or Consensus.