Who it is forA developer or founder who wants a working multi-file app built by an agent from a plain-language brief.
The problemStarting a new project from an empty folder takes days of scaffolding before any real feature exists.
The outcomeBenzi writes the frontend, backend and prompts of a new project in one chat session, with every write checked against the index.
What you need first
- Python 3.11 to 3.14 on Windows, macOS or Linux, for the pip install
- An email address for the one-time login code
- An API key from Anthropic or a non-Anthropic provider (DeepSeek, Groq, Kimi and others) for the Benzi agent, or an existing MCP-capable agent such as Claude Code or Cursor for the MCP server
Steps
- Install Benzi from the VS Code Marketplace.
- Click the Benzi icon in the left activity bar to open the Benzi: Settings panel.
- Pick your model, paste your API key, enter your email and press Send, then type the one-time code.
- Press Save model & keys, then Run Benzi.
- Describe the app you want in the chat and review each file it writes.
Prompt for your agent
In VS Code with Benzi running on an empty folder at <project path>, ask Benzi to build <short app description>. Expected output: a project with frontend, backend and any prompt files. Verify by running the app locally and checking that the main flow works end to end.
Caveat. The README's greenfield example is a single demo app, and results on other briefs will vary, so run and check the output yourself.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for oooscoos/Benzi and "Build a new app from scratch in one VS Code session"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer