grant-assistant-oss

Open-source AI co-pilot for nonprofits writing grant applications.

grant-assistant-oss is a developer-friendly toolkit and CLI that helps nonprofits, social enterprises, and CSR teams draft grant applications faster and better — using LLMs that understand donor-specific frameworks instead of producing generic text.

It is built and maintained by contributors from CSR Community, an initiative supporting nonprofit and social-impact organisations across Eastern Europe.

---

Why this exists

Funders increasingly use AI to screen incoming applications for alignment, evidence of impact, and methodological rigor. As a result:

• Generic "ChatGPT copy-paste" applications are systematically filtered out.
• Small NGOs without dedicated grant writers fall behind larger, well-resourced organisations.
• The same Theory-of-Change, logframe, and M&E plan get re-written from scratch for every donor.

Existing AI-grant-writing tools are closed-source, expensive, and English-only — exactly the audience-funding mismatch that public-interest software should fix.

grant-assistant-oss is the open alternative. It is:

• Donor-aware — built around the actual application schemas of EU Horizon, USAID, House of Europe, Google.org, EBRD, UN agencies and other public funders.
• Multi-language by default — English, Ukrainian, Polish, Russian.
• AI-assisted, not AI-replaced — humans stay in the loop; the tool produces structured drafts, not finished applications.
• Reusable — every donor template is a YAML file that the community can audit, fork, and improve.

---

Who is this for

Audience	What you get
Nonprofits & NGOs	Faster, donor-aligned drafts; reusable Theory-of-Change and logframe scaffolding.
Corporate CSR teams	The match command suggests grant programmes that fit a planned CSR initiative — useful when looking for co-funding or NGO partners.
Researchers & academic teams	Structured project narratives for science-policy and innovation grants.
Developers	A clean, hackable Python library you can embed in your own nonprofit tooling.

---

Quickstart

Status: alpha. The interfaces below are stable in intent but evolving in implementation.

Install (editable, from source):

git clone https://github.com/abondarev-guru/grant-assistant-oss.git
cd grant-assistant-oss
pip install -e .
export OPENAI_API_KEY=sk-...    # or ANTHROPIC_API_KEY

Draft a grant application from a short brief:

grant-assistant draft \
  --donor house-of-europe \
  --brief examples/brief-edu-initiative.md \
  --language uk \
  --output drafts/hoe-edu-2026.md

Match an existing CSR project to relevant funders:

grant-assistant match \
  --project examples/csr-project-mentorship.md \
  --top 5

Generate a stand-alone logframe / Theory of Change from a problem statement:

grant-assistant logframe \
  --problem "Rural youth in Eastern Europe lack access to digital skills training" \
  --output drafts/logframe.md

---

Architecture

                  ┌──────────────────────────┐
  brief.md ─────► │      Donor Adapter        │ ──► structured intent (JSON)
                  │  (eu-horizon, usaid, ...) │
                  └────────────┬─────────────┘
                               │
                               ▼
                  ┌──────────────────────────┐
                  │   Section Planners        │
                  │  problem · TOC · logframe │
                  │  budget · M&E · sustain.  │
                  └────────────┬─────────────┘
                               │
                               ▼
                  ┌──────────────────────────┐
                  │   LLM Refinement Layer    │ ◄── donor style guide
                  │   (OpenAI / Anthropic)    │ ◄── language pack
                  └────────────┬─────────────┘
                               │
                               ▼
                  ┌──────────────────────────┐
                  │      Output Composer      │ ──► .md / .docx / .pdf
                  └──────────────────────────┘

Each donor adapter is a YAML+Python pair describing:

• Required sections and word limits
• Eligibility checks
• Style and tone guide (extracted from public funder guidance)
• A scoring rubric mirroring the funder's published evaluation criteria

See src/grant_assistant/donors/ for the current adapters.

---

Roadmap

See ROADMAP.md for the full picture. Short version:

• v0.1 (current) — CLI scaffold, two donor adapters (House of Europe, generic EU project), English drafts.
• v0.2 — Ukrainian + Polish output, match command, logframe generator.
• v0.3 — score command (self-evaluation against donor rubric), 6+ donor adapters.
• v0.4 — Web UI for non-technical users, integration with Google Docs.
• v1.0 — Stable API, multi-donor batch mode, audit-trail export.

---

Contributing

We want contributors of three kinds:

1. Grant writers and program officers — even if you don't code, you can contribute donor templates and style guides (see docs/contributing-templates.md).
2. Python developers — pick up any good-first-issue ticket.
3. Nonprofits piloting the tool — open a feedback issue describing what worked and what did not.

Read CONTRIBUTING.md before opening a PR.

---

License

MIT. Use it, fork it, ship it. If your nonprofit benefits from this tool, an attribution back to the project helps other organisations find it.

---

Acknowledgements

• The CSR Community network of nonprofits and social-impact teams in Eastern Europe.
• The GitHub Social Impact Digital Public Goods guidance.
• All contributors who submit donor templates — these are the most valuable asset in the project.