CiteTrac builds fixed-fee document pipelines for small GovCon teams: automated requirement extraction where every "shall" traces to an exact source offset — with regression tests that stop it inventing a requirement or a past-performance claim. Your code. Your systems. Your data never leaves the building.
Get a free 20-minute matrix audit See a live teardownNumbers from Teardown #1 — a real, public SAM.gov solicitation. New teardown published weekly. Also see Teardown #2: what amendment churn does to a $200–600M pursuit.
Before anyone writes a win theme, somebody shreds the solicitation: Sections L, M, and C into a compliance matrix, by hand, in Word and Excel, under deadline. Industry surveys say the squeeze is getting worse — more bids, smaller teams, tighter turnarounds — while the shred process hasn't changed in twenty years. Generic AI tools promise to help, then invent a requirement that was never in the document. In federal proposals, that's not a bug. That's a disqualification — or worse.
A fabricated stat in a blog post is embarrassing. A fabricated requirement in a federal proposal can sink the bid. CiteTrac exists for the second case.
The pipeline shreds the solicitation into requirement statements — every "shall," "must," and "will," plus implied and pointer requirements. Each one is a word-for-word quote, because only exact quotes can be mechanically verified.
A grounding verifier re-locates every quote in the source at an exact character offset. Anything it can't trace is marked NEEDS SOURCE — displayed honestly, never guessed. Conflicts, ambiguities, and gaps get flagged with cited evidence.
A regression suite runs before any output ships; a fabricated claim aborts the run. Every defect the pipeline ever catches becomes a permanent test — the machine gets stricter with every solicitation it reads.
| Stage | Price | What you get |
|---|---|---|
| Matrix audit | Free | I run the pipeline on a public solicitation in your lane and walk you through the output for 20 minutes. No documents needed from your side. |
| Pipeline Blueprint | $1,000–$1,500 | Process teardown of how your team shreds today + a build-ready spec: architecture, guardrails, acceptance tests, fixed quote. 100% credited toward a build. |
| Build | $5,000–$12,000 | $5–7K single workflow (matrix extractor in your format); $9–12K full pre-bid pipeline (multi-format ingestion, amendment re-shred & diff, conflict reporting). You own all code, tests, and docs outright. |
| Maintenance | $400–$750/mo | Optional. Model updates, new solicitation-format handling, monthly accuracy report against a growing regression set. |
Compare: proposal-AI SaaS runs $15K–$60K+ per year, rented, generic, and cloud-hosted. A CiteTrac build is a one-time fee for a pipeline tuned to your matrix format and workflow — that keeps working after the last invoice.
Tell me a solicitation you're watching (or just your NAICS) and I'll run the pipeline on a public notice in your lane, then walk you through everything it caught — requirements, conflicts, ambiguities — live. You bring nothing but 20 minutes.