Spec-Driven Development for Cursor IDE — plan before you code, then let agents execute.
Quick Start • Everyday Commands • Pick a Workflow • Technical Docs »
git clone https://github.com/madebyaris/spec-kit-command-cursor.git
cd spec-kit-command-cursorThen open the folder in Cursor and type a command into chat. Start here:
/brief user-auth JWT authentication with login/logout
That's it — /brief plans the feature in ~30 minutes, then you implement it. For bigger work, see Pick a Workflow.
The ones you'll actually use day to day:
| Command | What it does |
|---|---|
/brief |
Quick 30-min plan for a feature (start here for most things) |
/plan |
Turn requirements into a technical plan |
/implement |
Build it, with progress tracking |
/audit |
Review the code against the spec |
/evolve |
Update the spec when things change mid-build |
Full command list, flags, and outputs are in the technical docs.
Just building a feature? (most of the time)
/brief my-feature → /implement my-feature
Complex or high-risk feature?
/research → /specify → /plan → /tasks → /implement
Whole app or big project?
/sdd-full-plan my-app → /execute-parallel my-app --until-finish
That's the 90% case. Everything else — deep research, parallel/cloud execution, heavy apps — is in the technical docs.
By default SDD is stateless and zero-setup. If you want agents to remember decisions and conventions across sessions, turn on memory:
/sdd-memory # pick a backend, or leave it off
Three choices: standard (default, nothing to set up), cursor-native (free Cursor Memories), or mem0 (free self-host). Details and trade-offs: technical docs.
- Technical documentation » — every command, subagents, skills, memory backends, cloud execution, architecture, and project layout
- Contributing — add commands, subagents, skills, and templates
- Report a bug · Suggest a feature
Thanks to ClavixDev for valuable ideas and suggestions!
MIT License — see LICENSE
Made with ❤️ by Aris
Try it: /brief hello-world Create a simple hello world feature