Makemyspec bridges the gap between your vision and your coding agent's interpretation
Makemyspec assumes the role of a product manager, a designer, an engineer and an architect to discuss the idea with you - in Q&A format - to bring your vision to life
No fixed questionnaire. The interview adapts to your idea — following threads, challenging vague answers, and stopping only when the picture is complete.
Every question comes with a concrete, reasoned recommendation. Accept it in one keystroke or override it. No "it depends" non-answers.
CLAUDE.md/Agent.md, PRD, user stories, ADRs, and a stack manifest — all generated in one pass, tailored to your chosen coding agent.
Whether you're using Claude Code, Cursor, Devin, Codex, or any other coding agent — Makemyspec formats the instruction files exactly as that agent expects it.
Four-layer security — structural detection, LLM classification, XML sandboxing, and output validation — keeps malicious inputs from poisoning your docs.
Paste your current workspace rules or custom agent instructions. Makemyspec incorporates them into the generated files without overriding anything you've set.
Not generic boilerplate. Documents built from your answers, in the exact format each agent expects.
Coding agents hallucinate when context is thin. Makemyspec front-loads the discovery work so your agent builds what you actually meant, not what it guessed.
No more opening a blank file and trying to remember what a good Agent.md looks like. Makemyspec produces it from your answers, not from your memory.
Makemyspec runs as a MCP tool inside Claude Code, Cursor, Codex or Devin — wherever you already work. No new app to open, no context to switch.
Every project gets the same rigorous interview regardless of how you phrase the idea. No project gets shortchanged because you were in a hurry.
Makemyspec plugs into your coding agent as an MCP server. Once connected, the whole flow is automatic.
Add Makemyspec to your coding agent's MCP config. One JSON snippet, one restart. Done for every future project.
Tell your agent what you want to build in plain language. Makemyspec detects the intent and starts the interview automatically.
Makemyspec interviews you across five stages: problem, scope, architecture, constraints, and agent selection. Each question has a recommended answer you can accept in one word.
Six documents land in your project root. Your agent reads them, understands your project fully, and starts building without asking for clarification.
Makemyspec generates agent instruction files in the exact format each tool expects — right filename, right location, right structure.
Claude Code
Cursor
Devin
Agentic Flow
Google Gemini
OpenAI Codex
"I used to spend an hour writing CLAUDE.md before any new project. Now Makemyspec does it in 15 minutes and it's more thorough than anything I was writing myself."
"The interview pushes back when I'm vague. It asked me three follow-up questions about my target user before I realised I'd been building for the wrong person."
"Switching from Claude Code to Cursor mid-project was painless — Makemyspec just regenerated the docs in Cursor's format. That alone was worth the subscription."
One free spec to try it. Then pay per project or go unlimited.
Run your first requirements interview free. No API key. No credit card. Just connect and start building.
Already have an account? Sign in · Questions? Read the docs