Local Development
Develop Runwork apps from your terminal with any AI tool
Install the Runwork CLI, clone or create an app, open it in Claude Code or Cursor, and start building. The platform creates CLAUDE.md + AGENTS.md on the fly during development, and SKILL.md is auto-generated for your app. Live preview without managing any infrastructure locally.
Powered By
The Challenge
- AI code assistants don't understand your platform or framework
- Local development requires Docker, databases, runtime dependencies
- Deploying from local to production requires CI/CD pipelines
- Switching between web editors and local editors means context switching
- AI-generated code needs platform-specific knowledge to work correctly
The Runwork Solution
- Platform creates CLAUDE.md + AGENTS.md on the fly during dev, SKILL.md auto-generated per app
- Remote sandboxes handle all execution, no local dependencies needed
- `runwork deploy` ships to production in one command
- Work on the web or locally, same apps, same workspace
- Knowledge files from the platform make AI output framework-aware
What You Can Build
Claude Code + Runwork
Clone app, run `runwork dev`, the platform provides CLAUDE.md + AGENTS.md live. Claude builds features that deploy correctly first try.
Cursor with Full Context
SKILL.md (auto-generated per app, also downloadable from dashboard) gives Cursor your app's entities, endpoints, workflows. AI suggestions are always framework-aware.
Zero-Config Live Preview
`runwork dev` boots a remote sandbox and streams your changes. See results instantly, no Docker or local runtime.
Team Handoff
Build an app locally with AI, deploy it, your team uses it on the web dashboard immediately.
Codex Integration
AGENTS.md provides context for OpenAI Codex. Same knowledge files, different AI tool.
Three Dev Tiers
CLI only (zero deps), add types with bun install, or full local frontend with Vite.
Git-Based Sync
Your files sync to remote sandboxes via standard git protocol. No proprietary sync tools.
A Day in the Life
9:00 AM - Setup. Install the CLI, run runwork init my-crm, open the project in Claude Code. Run runwork dev and the platform creates CLAUDE.md + AGENTS.md on the fly, plus SKILL.md for your app. The AI understands the entire framework before you type a single prompt.
9:30 AM - Building. Tell Claude "Add a customer entity with name, email, company, and a list view." Claude generates correct framework code because it has full context from the knowledge files. No guessing, no hallucinated APIs.
10:00 AM - Preview. Run runwork dev to boot a remote sandbox. A live preview opens in your browser. Make changes in your editor, see them instantly. No Docker, no database setup, no local runtime to manage.
11:00 AM - Expanding. Add integrations and workflows. The AI knows the framework patterns from SKILL.md, so it generates correct MCP server connections and automation triggers without you explaining the syntax.
12:00 PM - Ship it. Run runwork deploy. Your app is live. Invite your team. They open the dashboard and the app is there, ready to use. You built a production CRM in one morning with vibe-coding from your terminal.
Frequently Asked Questions
Do I need to install anything locally?
Which AI tools work with Runwork?
Can I switch between web and local development?
What languages and frameworks does it use?
Related Use Cases
Ready to build?
Start a 14-day free trial. No credit card required.