Open tooling, agent-run engineering, and the products built on both.

GrantPath

  • #grants
  • #consulting
  • #local-first
  • #explainable-scoring

Scope. GrantPath is a Back Road Creative product, spun out of the GrantBridge system design. The source is not public and there is nothing here to install. This page is what it does and who it is for. Asking about access is a conversation; the About page is where to start one.

The gap is not the grants

Grantees rarely lose because no grant existed. They lose in the gap between a listing in a database somewhere and an application that actually went out — hunting listings, re-typing the same organizational facts into every form, and starting each narrative from a blank page. That gap is where a consultant’s hours go, and hours are the only inventory a one-person practice has.

GrantPath is one system that closes it end to end: a consultant sits down with a grantee and leaves the session with applications moving, not a research list. Because most of it runs unattended, one operator can carry a book of clients instead of drowning in one.


Discovery to award

  • A catalog that maintains itself. Opportunities are retrieved from the federal Grants.gov API and from curated CSV packs chosen per client vertical. A listing that vanishes upstream closes itself with the reason recorded, and a funder-marked close is told apart from one that merely dropped out of the feed — so you never sit down in front of a grant that already closed, and “why is it gone” is on the record.
  • A fit score that shows its work. Every match is ranked 0–100 across five weighted dimensions — eligibility, geography, focus area, award size, and time to the deadline — and every point carries a stated reason. Where the listing is silent the score gives partial credit and says so, and a weak match names the one fact that would raise it. There is no black box to explain away at the table.
  • A profile that fills itself. An EIN lookup against the public nonprofit record brings back identity, location, revenue and capacity, and nothing is saved until the consultant has reviewed it. The rest is checkboxes and selects. Nothing is asked twice.
  • A checklist that writes itself. Shortlisting an opportunity creates the application with its checklist already built from the opportunity and the profile — the deadline task, the narrative, the budget, the letters of support, and the federal SAM/UEI registration item already marked done when the profile proves the registration. Required items gate submission.
  • Documents from facts already entered. The letter of intent, the boilerplate, the budget summary and the cover sheet fill themselves from what is on file and mark [ADD: …] exactly where a human is still needed. The software never invents a number or a claim it was not given — which is what makes the draft worth trusting.
  • One screen for the practice. A six-stage pipeline moves each application from shortlisted through preparing, submitted, and awarded or declined to closed, gated so a step cannot be skipped. A dashboard carries awarded-to-date, win rate, upcoming deadlines and per-client fresh matches; five reports export to CSV and Markdown; and each grantee gets a read-only portal into its own record.

Automatic, one-click, or deliberately human

The line between what runs itself and what waits for a person is drawn on purpose, and drawing it is what makes the automatic parts safe to leave alone.

It runs itself: nightly and quick syncs on cron or systemd, self-closing listings, a staleness banner if the timer stops firing, and a digest that mails only when there is something to say — including what is newly matched for each client, so an opportunity reaches the operator without anyone opening the dashboard.

One click: EIN prefill, shortlist-to-checklist, generate a document, export.

Deliberately human: portal submission stays manual, because the grantee’s own credentials are required for it. The [ADD: …] markers and the draft review gate keep a person in the loop wherever judgment belongs.


Where the data lives

Everything is in a local database file, exportable in full to Markdown and CSV at any time, and each grantee can pull reports scoped to its own organization. The core runs offline; syncing the catalog is the one explicit online step. The record outlives any vendor, including this one.


Who it is for

A consultant carrying several small nonprofits at once, who needs discovery, matching, preparation and tracking to be one system rather than four subscriptions and a spreadsheet — and who would rather explain a score than defend one.

Deep post-award compliance is deliberately not here. When a client wins an award that carries real reporting, drawdown and closeout obligations, that is the designed handoff to GrantBridge, the practice system this product was spun out of. The seam between the two is the design, not a missing feature.


Where it stands

The system runs today: discovery, matching, the application pipeline, document generation, the dashboard and the reports are all built and exercised by an automated suite on every change. The step that has not happened is the one that matters — the first sit-down with a real grantee. Until it does, version control here holds no release to date one from, and the card on the work ledger reads in development rather than carrying a month.

Nothing on this page is an announcement about availability. If it is interesting, the About page is the way to reach me.

← Back to Home