craft

An coding agent built with Bun, mostly on top of the Zen/Opencode API.

Give it a goal and watch it write code for you.

``js
// This is all an agent really is
// everything else is traditional data plumbing
You are an autonomous coding agent.
Goal: ${goal}
Context: ${context || 'None'}
So far: ${memory.results.join('; ') || 'Nothing'}
`




Install This Tool Now

`bash
bun install -g @skootsky/craft
`

Usage

`bash


Run the agent with a task

craft "write a CLI timer in Rust"

With context

craft "add tests" --context=/path/to/project
`

Setup

1. Copy .env.example to .env
2. Add at least one API key:
- API_KEY_ZEN for Open Code Zen

3. Set DEFAULT_PROVIDER env var to skip provider selection (optional)

Note: Multi model is very tempting, but also makes dialing in quality and taste very hard. I have attempted to split the difference by only doing OpenCode Zen models (which are a good collection!)

You might want to focus on Anthropic or OpenAI models from your favorite providers. That's up to you. Getting this right is an important part of building with AI, they aren't all the same, and not all providers are the same.

Commands and Local Development

| Command | Description |
|---------|-------------|
| craft "" | Run the agent with a task |
| bun run build | Build for distribution |
| bun run dev | Run from source (fast iteration) |
| bun run test | Run tests |

Project Structure

`
src/
agent.ts # Main CLI entry point
client.ts # LLM client wrapper
memory.ts # Persistent scratchpad for agent state
picker.ts # Provider selection
providers.ts # Available LLM providers
`

Features

- Native Bun parsing: Uses process.argv instead of commander
- Memory system: Tracks task, steps, and results across iterations
- Multi-provider support: OpenAI, Claude, and Zen integration
- Cost estimates for Zen (Opencode) models.
- Structured logging: Wide, canonical event logging throughout

This project was created using bun init` in bun v1.3.5. Bun is a fast all-in-one JavaScript runtime.