Darce

AI coding agent that lives in your terminal.
Reads, writes, edits code, runs commands, searches codebases.
One command to install. One command to start.

Website · npm · Get Started · Pricing · Dashboard

> fix the authentication bug in login.ts

  I'll read the file first.

  ○ Read src/auth/login.ts
    1  import { verify } from './jwt'
    ... 45 more lines

  Found it — token expiry compares seconds vs milliseconds.

  ● Edit src/auth/login.ts
    File updated

  ● Bash npm test
    24/24 tests passing

  Fixed. Wrapped the Unix timestamp in * 1000.

qwen3-coder · 3.1k tokens · $0.0008 · 6s

Why Darce?

Get Started

npm install -g darce-cli
darce login
darce

Requires Node.js 22 or newer.

That's it. No config files. No API keys to copy. No Docker.

What Can It Do?

Fix bugs — Describe the issue, Darce reads the code, finds the problem, fixes it, runs your tests.

Build features — "Add a dark mode toggle to the settings page" — Darce creates the files, writes the code, wires everything up.

Refactor — "Convert this class component to a hook" — reads the file, rewrites it, verifies nothing broke.

Explore codebases — "How does authentication work in this project?" — searches files, reads code, explains the architecture.

Run commands — "Install tailwind and set it up" — runs npm, creates config files, updates your code.

Features

/help     List commands          Ctrl+M   Switch models
/model    Pick / search models   Ctrl+C   Cancel / Exit
/clear    Reset conversation     Up/Down  Input history
/cost     Session costs          """      Multi-line mode
/compact  Shrink context

• 7 tools — Read, Write, Edit, Bash, Glob, Grep, WebFetch
• Smart routing — auto-picks the best model for each task
• Streaming — responses appear line-by-line as they generate
• Git-aware — knows your branch, changes, and recent commits
• Session resume — darce --resume picks up where you left off
• Context compaction — stays fast even in long conversations
• Cost tracking — real-time token count and spend in the status bar
• Account dashboard — usage stats at cli.darce.dev/dashboard

Models

Every tool-capable model on OpenRouter — 300+ and counting. The list is fetched live (cached for 24h), so new models show up the day they launch.

Open the picker with /model (or Ctrl+M) and type to search, or jump straight to one:

/model sonnet-5.5        # switch by name
/model kimi              # search — switches if there's one match
/model someone/new-model # any OpenRouter model ID
darce --model openai/gpt-5.6-sol

The picker shows context window, price per million tokens, and vision/reasoning support for each model. Default: qwen/qwen3-coder.

Pricing

Start free. Upgrade when you need more. Cancel anytime.

darce login           # Start free
darce upgrade         # Upgrade to Builder or Power

Or sign up at cli.darce.dev

Slash Commands

Config

darce login handles everything. For power users:

// ~/.darcerc
{
  "apiKey": "darce-...",
  "apiBase": "https://api.darce.dev",
  "router": {
    "default": "qwen/qwen3-coder",
    "rules": [
      { "when": "large-context", "use": "google/gemini-3.1-pro-preview" },
      { "when": "complex-reasoning", "use": "anthropic/claude-sonnet-5.5" }
    ]
  }
}

Contributing

git clone https://github.com/AmerSarhan/darce-cli.git
cd darce-cli
npm install
npm run dev           # Run from source
npx tsx test.ts       # 117 tests
npm run build         # Build for production

Star History

If Darce saved you time, drop a star. It helps others find it.

Built by darce.dev
_{MIT License}