Cake Proposal 🎂

Write stronger research proposals, faster — with an AI assistant that actually knows how grant writing works.

Cake Proposal is a VS Code extension built for one job: helping researchers turn an idea and a funding call into a polished, submission-ready proposal. It's made by ResearchCake.com for academic and research-grant proposals — not commercial bids or general-purpose writing.

Why not just use ChatGPT?

You can paste your ideas into a chatbot and get plausible-sounding text back. But a chat window doesn't know your funding call, can't read your files, invents citations, and forgets everything once the conversation gets long. You end up copy-pasting fragments between tabs and fact-checking every reference by hand.

Cake Proposal is different because it works where your proposal lives and does the research for real:

• It knows the craft. Successful proposals aren't just well-written — they're well-argued. The assistant follows a proven 8-phase workflow (understand the call → map the research landscape → build the argument → design the methodology → write → review like a skeptical reviewer → revise → export), and it knows what reviewers look for at each stage.
• Real literature, real citations. It searches actual academic databases (OpenAlex) and the web, reads the papers it finds — including PDFs — and manages your references in a bibliography file. Citations in your draft point to papers that exist.
• It works on your files, in your workspace. Your funding call PDF, your CV, your previous proposals, your draft — all right there. Attach any file with @mentions, and the assistant edits your draft directly instead of making you copy-paste.
• It plays the reviewer before the reviewers do. The critical-review phase scores your draft against the criteria real panels use — weak arguments, unsupported claims, feasibility gaps — so you fix problems before submission, not after rejection.
• From idea to Word document. When you're done, export straight to a formatted .docx (or PDF, PowerPoint, Excel) without leaving VS Code.
• Built for the long haul. Proposals take days or weeks, not one chat session. Sessions are saved and resumable, long conversations are compressed intelligently so you never hit a wall, and heavy research runs in the background so it doesn't clutter your workspace.

How it works

Tell the assistant about your funding call and your idea. It guides you through the phases of proposal development:

1. Project Context — Understand your funding call, team, and constraints
2. Research Intelligence — Analyze the academic landscape and identify the gap your work fills
3. Argument Architecture — Build the logical spine that makes a reviewer say "yes, this needs to be done"
4. Methodology & Evidence — Design a credible, feasible research plan with risks and contingencies
5. Narrative Writing — Turn the argument into polished prose tailored to your funder's format
6. Critical Review — Score the draft like a skeptical reviewer would
7. Revision — Fix the weaknesses and iterate until strong
8. Finalization — Export to Word (.docx) or other formats

You stay in control throughout: the assistant plans its work in a visible task list, you can review every change it makes to your files, and approval prompts (optional) let you decide before it writes files or runs commands.

What's under the hood

• Research tools — Academic search via OpenAlex (free, no API key), web search, deep literature reviews, URL fetching, and a controlled browser for JavaScript-heavy sites. A dedicated research assistant can run extended literature scans in the background and report back just the findings.
• Document handling — Reads PDF, DOCX, PPTX, and XLSX; academic PDFs are parsed into clean structured text (via GROBID) rather than raw dumps. Generates Word, PDF, PowerPoint, and Excel output.
• Reference management — Sources live in a references.json file; cite them in your draft with [@key] and the assistant keeps everything in sync.
• Precise editing — Surgical edits to your draft instead of wholesale rewrites, with safeguards against editing the wrong passage.
• Chat your way — Sidebar panel or full editor tab, session history, message branching (edit and resend past messages), and your choice of model: Claude, GPT, Gemini, Grok, and more.
• Cost-conscious by design — Long conversations are compressed with a cheap model, research results are kept out of your expensive context, and prompt caching is maximized. Long sessions stay affordable.
• Reliability — If your primary model has an outage, the assistant automatically falls back to a backup so your work isn't interrupted.

Getting Started

1. Install the extension from the VS Code Marketplace
2. Open a folder where you want to write your proposal
3. Click the Cake Proposal icon in the Activity Bar
4. Sign in with your ResearchCake.com account (usage-based billing, live balance display), or bring your own OpenRouter API key in Settings
5. Start chatting — tell the assistant about your funding call and research idea

Configuration

Open Settings → search for Cake Proposal. Most users only need the common settings below.

Common settings

Advanced settings

These point at ResearchCake's internal services. You normally won't need to change them — signed-in users get sensible defaults automatically.

Privacy

Your draft files stay on your machine. Only the prompts and the file content you choose to share are sent to the LLM provider you've selected (e.g. Anthropic, OpenAI, Google, xAI — routed via OpenRouter or your own API key).

Requirements

• VS Code 1.85.0 or later
• A ResearchCake.com account or an OpenRouter API key

Support

• Website: researchcake.com
• Email: support@researchcake.com

License

This is proprietary software. See LICENSE for details.

Copyright © 2025-2026 ResearchCake.com. All rights reserved.