Grantsville

An ED at a sub-$500K nonprofit who never wanted to write grants. They need them to survive, can't afford the $5,000 consultant, and can't afford to send a generic AI draft a funder will throw away in ten seconds. The Accidental Grant Writer.

The asset isn't missing — past proposals, financials, impact reports, board decks are all sitting in Google Drive folders. They've just never been connected to the next application.

Upload, classify, extract.

The org's documents go in. The vault classifies each file, extracts structured data across 27 document types, and surfaces a readiness gauge — what's filled, what's missing, what to chase down next. The vault has to populate itself, or it never gets populated.

Paste a guidelines PDF, get a brief.

The system pulls funder requirements, deadlines, evaluation criteria, and compliance rules out of the guidelines doc. You confirm the parsed brief before generation starts — wrong-on-paper beats wrong-in-a-draft.

Submission-ready proposals, grounded.

The agent pre-fetches the full vault context, maps funder requirements to your deliverables, builds budget narratives from your actual financials, and assembles language from proposals that already won. Every sentence has to be defensible against a citation in the vault.

Six-axis quality gate before you see it.

Every deliverable is scored on NOFO Coverage, Budget Coherence, Voice Match, Compliance, Funder Alignment, and Org Readiness — then it sits in an operator review gate before anything goes out the door. The weakest dimension has to be unskippable, or the gate is decoration.

Vault as moat, not the model.

The org’s documents are ingested into a queryable knowledge base — versioned, with funder-specific variants and a full audit trail. The model is a renderer over the vault, not the source of truth.

Document intelligence over manual tagging.

Uploads run through a multi-step pipeline that classifies the file (990, audit, proposal, board minutes…), routes it to the right extractor, and pulls the relevant atoms — all without asking the user to label anything.

Agent-native architecture.

AI capabilities are markdown prompt files plus a small set of tools — read/write vault, process documents, draft sections, manage workflow state. There are no hardcoded generation chains.

Per-deliverable pricing, not per-seat access.

$99 per Letter of Inquiry, with Stripe at the end of the workflow. First project is free. There is no monthly access fee — you pay when the vault produces something you ship.

Proposal acceptance rate — the only metric tied to a real user outcome, and the one that tells you whether the vault thesis actually held in production.

Payments and multi-tenancy on Stripe. Document classification across 27 types. The agent on markdown prompt files. The six-axis review gate with per-axis sign-off. First paying customers walking through the four-surface flow end-to-end.