OpenGrants OS

The open-source Autonomous Agentic AI MCP Server for grants, public-goods funding, and open-source AI sustainability.

Eliminate the discovery tax. Treat traditional grants and donations / sponsorships / quadratic funding as equal first-class citizens so open-source AI and public-goods projects can compete on a level playing field.

Try it now

Launch the interactive shell →

A complete, offline-capable demo: Discover, Constellation (3D knowledge graph), Pipeline, Draft, Console, and Profile — all in one page. No install required.

Or open the file directly: demo/index.html

Why OpenGrants OS?

Most funding tools are either closed SaaS or fragmented portals. OpenGrants OS is:

• A first-class MCP Server any agent (Claude, Cursor, custom agents) can call
• An autonomous multi-agent system for discovery → ranking → drafting → evaluation
• Explicitly optimized for open-source AI, open weights, and public-goods funding
• Fully open-source (Apache 2.0), self-hostable, and provenance-first
• Already dogfooding itself (see docs/DOGFOODING_OPENGRANTS.md)

Features (v0.2.0)

Capability	Status
Live Grants.gov search	✅
NSF Awards search	✅
Public-goods / donation sources (GitHub Sponsors, Open Collective, Gitcoin, OpenSSF, NumFOCUS, PSF, NLnet, Prototype Fund…)	✅
Profile Steward	✅
Grounded proposal drafting (grants + sponsorship vehicles)	✅
Continuous ingestion + evaluation harness	✅
Multi-agent ranking with open-source bias	✅
Interactive shell + 3D Constellation	✅
FUNDING.yml ready	✅

Quick Start (MCP Server)

git clone https://github.com/ANAMIZED/opengrants-os.git
cd opengrants-os
pip install -e .
python -m opengrants          # starts the MCP server (stdio)

MCP Client Config (Claude Desktop / Cursor)

{
  "mcpServers": {
    "opengrants": {
      "command": "python",
      "args": ["-m", "opengrants"],
      "cwd": "/absolute/path/to/opengrants-os"
    }
  }
}

MCP Tools

Tool	Description
search_grants	Live Grants.gov keyword search with full provenance
get_grant_details	Detail view for a specific opportunity
list_open_source_relevant_grants	Open-source / AI biased discovery
list_public_goods_funding	Donations, sponsorships, quadratic & open-source funds
search_nsf_awards	NSF Awards public API
refresh_corpus	Continuous ingestion into local corpus
run_evaluation	Discovery + open-source relevance harness
upsert_profile	Create / update project or researcher profile
draft_proposal_outline	Grounded outline + short pitch (works for grants and public-goods vehicles)

Architecture

MCP Interface
    ↓
Discovery (Grants.gov · NSF · Public-Goods Catalog)
    ↓
Profile Steward → Ranking Agents → Drafting Agents
    ↓
Evaluation & Continuous Corpus

The interactive Constellation view maps every pillar, tool, source, and opportunity as stars orbiting OpenGrants OS.

Support the Project

• GitHub Sponsors (enable on the account → Sponsor button appears)
• Open Collective (create a collective and update .github/FUNDING.yml)
• See docs/DOGFOODING_OPENGRANTS.md for the full self-generated proposal outline

Development

pip install -e ".[dev,agents]"
ruff check .
python -m opengrants.evaluation.harness

Publishing & Discovery

• Official MCP Registry: server.json present → mcp-publisher publish
• Glama: glama.json present
• PyPI: ready (hatch build && twine upload dist/*)
• Demo: GitHub Pages from /demo

License

Apache License 2.0 — see LICENSE.

---

Built for the open-source AI community.
Make the funding landscape as open as the code we write.

https://github.com/ANAMIZED/opengrants-os