A full-lifecycle grant writing engine for Claude.
From finding the right opportunity to a submission-ready application — discovery, eligibility triage, cited research, drafting, budgets, and a mock-scorer review, all in your organization's voice.
Grant writing is slow, repetitive, and unforgiving: the same organizational boilerplate rewritten for every funder, requirements scattered across 60-page NOFOs, budgets that must reconcile to the dollar, and reviewers who score against rubrics you never see.
GrantForge streamlines that entire process for the people doing the work — nonprofits, schools, universities, local governments, and companies pursuing federal, state, foundation, and corporate funding. Every draft passes through a built-in house style that removes AI writing patterns, and every application is graded by a mock reviewer before it goes out.
- How It Works
- The Workflow
- Deliverables
- Installation
- Getting Started
- Components
- Repository Layout
- Responsible Use
- Contributing
- License
GrantForge learns your organization once, then reuses that knowledge on every application:
- Set up once. A guided interview builds your organization's dossier — identity and registrations (EIN, UEI, SAM), mission and programs, capacity and leadership, financials, past awards, and impact metrics. Upload past applications and GrantForge extracts answers from them into a reusable answer library, and learns your writing voice from your samples.
- Work the pipeline. Everything lives in a
.grantforge/folder that persists across sessions. Each completed application makes the next one faster — winning language is harvested back into your answer library.
Run the whole lifecycle with /grantforge, or invoke any stage directly by just asking:
| Stage | Say | What Happens |
|---|---|---|
| Discover | "find grants for us" | Searches Grants.gov, state portals, foundations, and corporate givers; returns a ranked, verified shortlist |
| Intake | "respond to this RFP" | Parses the announcement into a requirements matrix; eligibility check; go/no-go recommendation |
| Research | "research this grant" | Builds a cited evidence base (need data, intervention evidence) and funder intelligence (990s, giving history, priorities) |
| Draft | "draft the application" | Writes each section against the funder's verbatim prompts, within limits, in your voice |
| Budget | "build the budget" | Excel budget workbook plus a matching budget narrative, with verification math |
| Review | "are we ready to submit?" | Compliance audit, mock-scorer grading against the rubric, and a requirements-to-response compliance matrix |
| Track | "what's in our pipeline?" | Deadlines, statuses, win rate; harvests winning language back into your answer library |
- Word documents formatted to the funder's rules
- Portal-ready plain-text sections with character counts
- Excel budget workbooks with a matching budget narrative
- A compliance matrix proving every requirement was addressed
No API keys, accounts, or MCP servers required — research uses built-in web search.
claude plugin marketplace add ddanntheman/GrantForge
claude plugin install grantforge@grantforge-marketplaceOr try it for a single session without installing:
git clone https://github.com/ddanntheman/GrantForge.git
claude --plugin-dir ./GrantForge/plugins/grantforgeDownload dist/grantforge.plugin, then open it in a Cowork chat and click Save plugin (or install via Settings → Capabilities → Plugins).
plugins/grantforge/ is a standard Claude plugin directory (.claude-plugin/plugin.json, skills/, commands/). Place it wherever your environment loads plugins from. The nine skills are self-contained markdown and work individually if the target only supports skills.
Connect a folder where GrantForge should live (your grants folder), then say "set up GrantForge". Setup takes one guided conversation; uploading past applications and reports speeds it up considerably. After that, start anywhere in the workflow — paste an RFP, ask for discovery, or run /grantforge.
| Component | Purpose |
|---|---|
/grantforge |
Lifecycle orchestrator — runs the full workflow end to end |
grant-setup |
First-run interview; builds the org profile, answer library, and voice profile |
grant-discover |
Finds and ranks open funding opportunities matching your profile |
grant-intake |
Parses RFPs/NOFOs into a requirements matrix; go/no-go recommendation |
grant-research |
Cited evidence base and funder intelligence (990s, giving history) |
grant-draft |
Section-by-section drafting against the funder's verbatim prompts |
grant-budget |
Budget workbook and budget narrative with verification math |
grant-review |
Compliance audit, mock-scorer grading, compliance matrix |
grant-tracker |
Cross-session pipeline: deadlines, statuses, win rate |
house-style |
Shared writing standard; strips AI writing patterns from all prose |
.claude-plugin/marketplace.json Claude plugin marketplace manifest
plugins/grantforge/ The GrantForge plugin (source of truth)
.claude-plugin/plugin.json Plugin manifest
commands/grantforge.md /grantforge lifecycle orchestrator
skills/ The nine skills
dist/grantforge.plugin One-click installable for Cowork / Claude desktop
GrantForge drafts, researches, and checks — it does not submit. Every application still needs a human who knows the organization to verify facts, figures, and commitments before it goes to a funder. Treat the mock-scorer review as a rehearsal, not a guarantee.
Issues and pull requests are welcome. If GrantForge helped you win (or lose) a grant, feedback on what worked and what didn't is especially valuable — file an issue describing the funder type and where the output fell short.
Released under the MIT License.
Copyright © 2026 Drew Danner