Skip to content

Repository files navigation

GrantOps

English | Español | Português (Brasil) | 한국어 | 日本語 | Русский | 简体中文 | 繁體中文

GrantOps — Multi-Agent Grant Acquisition System

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.

Claude Code Gemini CLI Groq Node.js Playwright MIT


GrantOps Demo

£10M+ opportunities evaluated · 50+ autonomous extractions · 1 mission-critical engine

What Is This

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.

Features

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

Quick Start

# 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-url

The 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.

How It Works

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

Project Structure

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

About the Author: Ashish Sandhu

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.

I am O-1A Visa eligible (no H-1B lottery required), ensuring a frictionless sponsorship process for elite talent.

License & Trademark

The code is licensed under MIT. The "GrantOps" name and brand are governed by the Trademark Policy.

About

No description, website, or topics provided.

Resources

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages