OpenGrants OS

The ultimate open-source Autonomous Agentic AI MCP Server for grants discovery, matching, drafting, and lifecycle management — built for open-source AI and public goods.

Vision: Eliminate the discovery tax that disadvantages open-source AI builders. Provide transparent, grounded, autonomous infrastructure so public-goods work can compete for funding on equal footing with closed labs — including traditional grants and donation / sponsorship / quadratic public-goods funding.

What’s New in 0.2.0

• Public Goods / Donation sources: GitHub Sponsors, Open Collective, Gitcoin, Sentient, OpenSSF, NumFOCUS, PSF, and more as first-class funding opportunities

• NSF Awards search

• Profile Steward: create and maintain project/researcher profiles for personalized matching and drafting

• Drafting agents: grounded proposal outlines + short pitches for both grants and donation vehicles

• Continuous ingestion + evaluation harness (from 0.1.x)

• Multi-agent ranking with open-source bias

Related MCP server: Grants.gov MCP Server

Quick Start

git clone https://github.com/ANAMIZED/opengrants-os.git
cd opengrants-os
pip install -e .
python -m opengrants

MCP Client Config (Claude Desktop / Cursor)

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

MCP Tools

Resources

• opengrants://status

• opengrants://sources

Why Donation / Gifting Sources?

Many of the most important open-source AI contributions are funded outside traditional government grants — through GitHub Sponsors, Open Collective, Gitcoin quadratic rounds, foundation open-source programs, and direct public-goods funding. OpenGrants OS treats these as first-class citizens alongside Grants.gov and agency programs so that maintainers of public-goods AI can discover and apply for the full spectrum of support.

Architecture

MCP Interface → Discovery (Grants.gov + NSF + Public Goods) → Profile Steward → Ranking Agents → Drafting Agents → Evaluation & Corpus

Development

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

Publishing

• PyPI: Package is ready (hatchling). Publish with hatch build && twine upload dist/* once credentials are available.

• Official MCP Registry: server.json is present. Use mcp-publisher login github && mcp-publisher publish.

• Glama: glama.json is present; claim the listing after indexing.

License

Apache License 2.0

Built for the open-source AI community.
Let’s make the funding landscape as open as the code we write.

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