Your AI coding tool has no idea what you're building

Unless you tell it. That's the problem — and the fix takes 60 seconds.

Two reasons AI drifts off course

First: no project context. Copilot, Cursor, Claude Code, Windsurf — they all start every session knowing nothing about your project. Without your stack, conventions, and architecture, they guess. Wrong frameworks. Inconsistent patterns. Code that looks right but isn't.

Second: knowledge cutoff. Every AI model has a training cutoff date. That means it might suggest a library that has since changed its API, recommend a pattern the community has moved away from, or not know about a tool you're actively using. The model doesn't know what it doesn't know.

Both problems have the same fix: a context file that tells your AI exactly what you're using, how your project is structured, and what matters before it writes a single line of code.

What a context file does

Every major AI coding tool can read a context file at the start of every session. The file contains your tech stack, architecture decisions, coding conventions, project structure, and anything else that keeps the AI aligned with what you're actually building.

Think of it as onboarding for your AI — the briefing a new developer would need on day one: what's this built with, how is it structured, what are the rules. Except your AI reads it every single time.

SpecifyThat doesn't generate a spec or a PRD. It generates that file — the context your tool needs before you start prompting.

How SpecifyThat works

1

Describe your project

A sentence or two is enough. Paste a doc if you have one. AI uses this to understand the shape of what you're building.

2

AI figures out the rest

Behind the scenes, AI works through 13 strategic questions — stack, architecture, users, constraints, non-goals. You get a review screen with everything pre-filled. Adjust anything that's off. Most people change nothing.

3

Get your file

SpecifyThat compiles the answers into a structured context file. Copy it, download it, drop it into your project. Your AI tool is now briefed.

Where to put the file

The content SpecifyThat generates works for any AI coding tool. The only difference is the filename and where it lives. Here's where each tool looks:

GitHub Copilot.github/copilot-instructions.mdauto-loaded
Cursor.cursor/rules/context.mdcauto-loaded (AGENTS.md also works)
Claude CodeCLAUDE.mdproject root, auto-loaded
Windsurf.windsurfrulesauto-loaded
Bolt / Lovable / v0Project context fieldpaste on project creation
Any other toolAttach or pasteinclude at the start of your session

For auto-loaded files, your tool picks it up without any configuration. For tools like Bolt and Lovable, paste the content into the project context field when you start a new project.

Under the hood

SpecifyThat runs entirely in your browser. Your answers are sent to our server only to generate AI responses — we never store, log, or inspect your content.

  • API routes are thin proxies: validate rate limits → forward to AI → return response
  • Your generated files save to localStorage — never our servers
  • No accounts, no cookies, no tracking pixels beyond basic analytics
  • Free to use — rate limits apply to keep the service available for everyone

Ready to try it?

Start building

Free · No signup · Under 60 seconds

Feedback