Your past proposals become a cited library. New responses are drafted from it, in your voice, with every claim traceable and every gap flagged instead of invented.
Proposal Kit is self-hosted software that reads your firm's past winning proposals, extracts the reusable content into a cited block library (your genome), and drafts complete responses to new RFPs from that library in your firm's voice. Every claim traces to one of your documents; anything the library can't support is flagged NEEDS-INPUT rather than made up. A compliance map shows where each RFP requirement is answered, and a revision loop turns your edits into standing style rules. Sold by A2 Digital as a one-time install; you use your own AI API key.
Generic AI drafting fabricates placements, dates and credentials. Proposal Kit only writes what your library supports and flags the rest for a human.
The compliance map lists every RFP requirement and the draft section that answers it. Reviewers see coverage before they read a word.
Edits in review become style rules the next draft follows. The tenth proposal sounds more like you than the first.
Feed in past proposals (PDF or Word). Reusable blocks are extracted with source, date and variant tags; a coverage report shows what's strong and what's thin.
Requirements are parsed into a checklist. Each is matched to candidate blocks from the genome.
A full response is drafted in your voice. The compliance map links requirements to sections; unsupported items are marked NEEDS-INPUT.
Comment, revise, approve. Edits feed the style-rules store; the final renders to a branded document.
Prices in USD. Requires a Linux server or VM, Python 3.11+, SQLite, and an API key from your chosen AI provider (usage billed to you). Onboarding of the first genome is self-serve with the extraction script and guide.
Those are hosted content-library platforms priced per seat, built for enterprise proposal teams. Proposal Kit is one install, self-hosted, your own AI key, and built around a grounding rule: the draft cannot say anything your library doesn't support.
Your past proposals as PDF or Word. The more wins you feed it, the richer the library. The coverage report shows which topics are thin so you know where NEEDS-INPUT flags will cluster.
Your choice; you supply the API key. The reference install uses a frontier model for drafting and a cheaper one for extraction. Model usage is billed by your provider, typically a few dollars per proposal.
Only the prompts sent to your chosen AI provider under your own key, subject to that provider's terms. The library, drafts and style rules stay on your server.
Setup is self-serve with the extraction script and a written guide. Email support covers the software. We don't curate your library for you; that's the line between a product and a service.
No. It drafts and renders. A person reviews, approves and submits every response.
Bring three past wins and one open RFP. Thirty minutes.
Request a demo