English | Español | Português (Brasil) | 한국어 | 日本語 | Русский | 简体中文 | 繁體中文
Most grant search tools are passive databases. They list. They wait. You work.
Companies use AI to filter applicants. I just gave founders AI to choose capital.
Now it's open source. Local-first. Database-free.
£10M+ opportunities evaluated · 50+ autonomous extractions · 1 mission-critical engine
GrantOps turns any AI coding CLI into a full capital acquisition command center. Instead of manually tracking UK grant opportunities in a spreadsheet, you get a local-first, AI-powered pipeline that:
- Evaluates capital with a structured A-F scoring system (10 weighted dimensions)
- Strips guidance noise -- automatically extracts questions and word limits from 40-page PDFs
- Scans portals automatically (Innovate UK, GOV.UK, Swoop, 360Giving)
- Maps narratives -- aligns your "Infrastructure-First" startup proof-points to funder priorities
- Tracks everything in a single source of truth (
data/tracker.tsv) with integrity checks
Important: This is NOT a generic application generator. GrantOps is a filter -- it helps you find the few capital sources worth your time out of thousands. The system strongly recommends against applying to anything scoring below 70/100. Your time is leverage; spend it wisely.
Built by an AI Infrastructure Engineer who treats capital acquisition as a compiler for high-level logic. Read the full case study.
| Feature | Description |
|---|---|
| Auto-Pipeline | Paste a URL, get a full evaluation + report + tracker entry |
| A-F Evaluation | Match, gaps, strategic alignment, and "Win Theme" discovery |
| Narrative Mapping | Maps startup proof-points (metrics, awards) to funder requirements |
| Database-Free | Uses local TSV and Markdown. No cloud dependency required |
| Portal Scanner | Scans major UK grant portals directly from the terminal |
| Agentic Modes | 8 skill modes (grant, scan, draft) that define deterministic AI behavior |
| Human-in-the-Loop | AI evaluates and recommends, you decide and act. Always have the final call |
# 1. Clone and install
git clone https://github.com/ashishsandhu/grantops.git
cd grantops && npm install
# 2. Check setup
npx playwright install chromium # Required for portal scraping
# 3. Configure
cp config/profile.example.yml config/profile.yml # Edit with your startup details
# 4. Start using
# Run inside an AI CLI (Claude Code, Gemini CLI, or OpenCode)
/grantops https://www.gov.uk/grant-example-urlThe system is designed to be customized by the AI itself. Modes, archetypes, and scoring weights -- just ask your AI agent to change them. It reads the same files it uses, so it knows exactly what to edit.
You paste a grant URL or description
│
▼
┌──────────────────┐
│ Match Scorer │ Classifies: R&D / Innovation / Scaling / Regional
│ Detection │
└────────┬─────────┘
│
┌────────▼─────────┐
│ A-F Evaluation │ Eligibility red-teaming, strategic fit, word counts
│ (reads profile) │
└────────┬─────────┘
│
┌────┼────┐
▼ ▼ ▼
Report Draft Tracker
.md .md .tsv
grantops/
├── AGENTS.md # Agent instructions for slash commands
├── config/
│ └── profile.yml # Your startup identity and proof points
├── modes/ # 8 agentic skill modes
├── data/ # Your tracking data (TSV)
├── reports/ # Evaluation reports (Markdown)
├── jds/ # Raw grant descriptions
├── templates/ # Scanner and narrative templates
└── docs/ # Architecture and setup guides
I am an AI Infrastructure Engineer. I treat AI as Pure Leverage, using LLMs as "recursive thinking partners" to solve edge cases in complex systems.
- Experience: Building since 2017 (GPT-2/BART era).
- Scale: Architected systems for 10M+ MAU.
- Philosophy: Infrastructure-First, High-Leverage, Zero-Fluff.
The code is licensed under MIT. The "GrantOps" name and brand are governed by the Trademark Policy.