Real-time guidance
Writers receive practical suggestions while developing an application, grounded in the funder's criteria.
Overview
Guidance grounded in the funder's own instructions
CanGrant is an AI-enabled grant-writing tool built inside the AI Accelerator Hub, an Algonquin College Applied Research initiative in Ottawa. It gives writers real-time guidance while an application is being drafted rather than a verdict once it is finished, and the funder's own instructions supply the grounded truth for that guidance.
Behind the guidance sits a custom AI knowledge base of Canadian grant applications that were funded and applications that were rejected. CanGrant derives winning patterns from that evidence, so the advice a writer sees reflects what has actually persuaded a review panel rather than what a general-purpose model believes a grant should sound like. The agents work from the funder's instructions and that knowledge base, and they do not invent facts.
Outputs are formatted to meet funder standards. CanGrant was built by researchers for researchers, and it currently runs as an internal tool inside the Hub rather than as a public product.
Highlights
Grounded guidance for stronger proposals
Learns from what wins
A custom knowledge base distils patterns from successful and rejected Canadian grant applications.
Built by researchers
Outputs follow funder standards and reflect the needs of real applied-research teams.
Inside the product
What CanGrant looks like
Start a proposal, or ask CanGrant
The home screen, shown at the top of this page, opens on two ways in: start a proposal and work through it with the assistant, or ask CanGrant a question about a programme you are considering. Both draw on the same grounding — NSERC's evaluation criteria, applications that were funded, and the mistakes that got other applications rejected.
Grant Finder
Before anything gets written, an applicant has to be pointed at the right programme. Grant Finder captures the shape of the organisation and the work — legal type, province, focus areas, communities served, a project summary and past grants — and returns the Canadian grants that applicant is positioned to win.
The winning-pattern knowledge base
The patterns CanGrant reasons from are visible rather than hidden inside a prompt. The knowledge base holds 43 winning patterns distilled from funded CCSIF proposals, 15 of them marked critical, with the rest labelled important or helpful so a writer can tell which advice is load-bearing and which is polish.
Technology
Secure, research-grounded AI
AWS Bedrock provides the managed foundation models that power the reasoning, and Claude Sonnet drafts the grant content. The interface is a React application built on Next.js, sign-in and account security run through Google Sign-In, and the cloud infrastructure is provisioned with Terraform so the environment can be rebuilt from source rather than clicked together by hand.
- AWS Bedrock
- Claude Sonnet
- React
- Next.js
- Terraform
- Google Sign-In
Who builds it
Built by the Hub's student team
CanGrant is designed, built and maintained by Algonquin College students on paid research internships and co-op placements, running on enterprise-grade AWS infrastructure. The AI does the research, the drafting and the summarizing; the team reviews every output before it ships. Meet the team behind it, or read more about the AI Accelerator Hub.
More from the Hub
The other two products
- Vibe ConsultSector-tailored AI adoption strategy, reviewed by a person before it reaches you. Coming soon.
- VibeNewsDaily AI news for Ottawa, with summaries, a weekly podcast and a newsletter. Live now.
- All productsThe three AI tools built by the AI Accelerator Hub in Ottawa.