Cairn Federal went live this summer at cairnfed.ai. It's a capture platform for small businesses chasing U.S. federal contracts and grants: it scans SAM.gov, Grants.gov, and USAspending, scores every notice against your company's actual capabilities, and carries the ones worth chasing all the way to a submitted proposal. I started building it in June. Here's what it does and how it's put together.
The problem it solves
If you've pursued federal work at a small company, you know the routine. Somebody scrolls SAM.gov by hand. Opportunities live in a spreadsheet. The solicitation is a dense PDF with requirements scattered through it, and the deadline math happens in someone's head. Cairn Federal is built for the people carrying that load: capture managers, proposal writers, and business-development leads at small and mid-size businesses. It covers contracts and grants both, so a nonprofit or university pursuing Grants.gov funding gets the same pipeline discipline as a contractor. The Starter tier is free, which means you can find out whether it fits before paying anything.
From first notice to submitted proposal
It starts with search. Every SAM.gov and Grants.gov notice gets a fit score from 0 to 100 against your NAICS codes, set-aside certifications, target agencies, and capability narrative, with the reasons spelled out rather than hidden inside a model. When a high-scoring notice requires a certification you don't hold, it's flagged ineligible instead of quietly ranked, so the tool never talks you into a bid you can't legally win. Setting up the profile behind those scores takes one UEI: Cairn Federal pulls your SAM.gov registration, CAGE code, NAICS codes, and set-asides for you.
From there, pursuits move through a kanban pipeline with bid/no-bid gate reviews and an explainable win-probability score. My favorite feature sits a step before the notice even exists. Most federal work is already being done by somebody, so Recompete Radar sweeps USAspending every week for awards in your space that are approaching the end of their term, names the incumbent, and scores the relevance. You get to position for the follow-on before the solicitation drops.
The proposal workspace handles the back half: versioned sections where AI drafts are clearly tagged as AI, a compliance matrix, color-team review rounds, and a submission package that bundles the proposal as Word and PDF with the checklist and cover sheet. A readiness check blocks the package until open compliance items and must-fix comments are resolved. On the Enterprise tier, grant applications file electronically through Grants.gov's system-to-system interface. There's an iPhone app too, included with every account.
The AI drafts. You approve.
The AI runs on LangGraph workflows: graph definitions authored in Cairn Federal's own source code and executed by the Wildwood platform's LangChain and LangGraph engine. The showpiece is the solicitation shred. Upload the RFP or funding notice, and a workflow splits it into chunks, extracts every requirement in parallel, works out whether it's reading a contract or a grant, and proposes a section outline. Then it stops. A person reviews that outline (approve, edit, or reject) before a single word of proposal gets written, and the whole run streams through the graph on screen, node by node.
Other workflows draft individual sections grounded in your approved past performance and content library, run the weekly recompete sweep, and pull award history from USAspending into bid/no-bid scorecards as suggested justifications waiting for your sign-off. There's also a chat assistant with 44 tools that can read and act across the whole app. The rule for all of it is the same, and it's enforced in code: reading is free, but any action that changes your data pauses, lists exactly what it's about to do, and waits for a yes.
Your documents do two jobs
The proposal writer's grounding comes from documents you control, and every document you attach to a pursuit gets one of two roles. A content source is factual grounding: the AI draws capabilities, evidence, and substance from it. A style guide is the opposite: the AI matches its tone, structure, and formatting but is told never to treat it as a source of facts.
On top of that sits the voice profile. Upload a few past proposals once, and Cairn Federal distills them into a compact profile of how your company writes: tone, sentence rhythm, structural habits, preferred terms, and what to avoid. Drafts mirror that profile, and it's style only by design; the AI never copies facts or client names out of your samples. Even the AI review rounds know the profile exists, so they don't flag the very phrasing the writer was told to adopt.
How it's built
Under the hood it's a .NET 10 API, a React web app, and a React Native mobile app sharing one framework-free domain library, so the search and pipeline logic on your phone is literally the same code as on the web. It runs on the self-hosted Kubernetes cluster I wrote about in my last post. And it's built on Wildwood, my backend-as-a-service platform: authentication with passkeys, subscriptions and payments, feature entitlements, document storage, retrieval collections, notifications, and the AI gateway and workflow engine all came from the platform. That's the main reason a product this size could go from first commit in June to a live deployment in July.
If you pursue federal contracts or grants, try it: the Starter tier is free, and the fit-ranked search alone will tell you whether it earns a place in your week. And if you're wondering what it takes to get an app like this built for your own business, that's the work I do at Kuhn Consulting: idea to live, hosted software. Let's talk.