dboudreau00/countersign-rfp-rfi-rfq-studio-for-ai-assisted-workflows

Countersign is a single HTML file that helps a team answer SaaS RFPs, RFIs, and RFQs. It works by first building a bucket of your own documentation, up to 500 sources, then reading an incoming RFP file in PDF, Word, Excel, CSV, or pasted text form and pulling out every question inside it. For each question it generates an answer grounded in your documentation, along with citations back to the source material and a score showing how well each answer is covered. Finished answers can be exported as a Word document, spreadsheet, or CSV file, or written back into the original spreadsheet the issuer sent. The whole app is one HTML file that can be uploaded to any web host with no server code required. Each visitor's work is saved in their own browser storage, and each person picks an AI provider in the settings panel, either Claude with their own Anthropic key, Google Gemini, or any OpenAI compatible endpoint. A team can also export a shared workspace file containing the documentation bucket and question history, upload it next to the HTML page, and every new visitor will automatically load that same starting point. Rather than have every team member paste their own API key, the project includes two optional server-side proxies, one written in PHP for ordinary shared web hosting and one in Node.js for a VPS, that hold a single API key on the server so browsers never see it. Both proxies restrict which AI models can be used, cap how long a single response can be, and can require a shared team password so a stranger who finds the page cannot run up the bill. The README is careful to warn that an API key should never be typed directly into the HTML page since anyone can view the page's source, that the shared workspace file is publicly readable wherever it is hosted, and that the bundled proxies limit request size but do not limit how often requests can be made, so a real deployment facing the internet should add its own rate limiting. A companion script called engine.js goes further, loading the documentation bucket on the server itself and exposing a question answering endpoint that other tools, bots, or scripts can call directly.