craft-code 0.1.0

A Claude Code-like AI coding assistant built with LangChain Deep Agents

Code Craft

A Claude Code-like AI coding assistant built with LangChain Deep Agents.

Features

• File system access — reads, writes, and edits project files
• Shell execution — runs commands (tests, linters, build tools)
• Agentic planning — breaks tasks into steps with write_todos
• Human-in-the-loop — asks approval before destructive operations
• Persistent memory — remembers project context across sessions
• Subagent delegation — spawns code-reviewer and researcher subagents
• Skill system — lazy-loaded workflows for testing, code review, etc.

Quick Start

# 1. Clone and install
cd code-craft
uv venv && source .venv/bin/activate
uv pip install -e .

# 2. Set your API key
cp .env.example .env
# Edit .env and add your ANTHROPIC_API_KEY

# 3. Run
coding-agent

Usage

# Interactive CLI
coding-agent

# Or run directly
python -m code_craft

Inside the CLI:

• Type your coding request
• The agent will ask for approval before file writes and shell commands
• Use /new to start a fresh thread, /quit to exit

Programmatic Usage

from code_craft.agent import build_agent
from code_craft.config import AgentConfig

config = AgentConfig(project_root="/path/to/your/project")
agent, checkpointer, store = build_agent(config)

Configuration

Set these in .env or as environment variables:

Variable	Default	Description
ANTHROPIC_API_KEY	(required)	Your Anthropic API key
MODEL	anthropic:claude-sonnet-4-6	Model to use
PROJECT_ROOT	.	Root directory for file operations

Architecture

code_craft/
  agent.py              — Agent factory with backends, subagents, skills
  cli.py                — Interactive Rich-powered CLI
  config.py             — AgentConfig dataclass
  interrupt_handler.py  — Human-in-the-loop approval UI
  prompts.py            — System prompts
  tools.py              — Custom tools (git, delete, tests)
skills/
  testing/SKILL.md      — Test writing workflow
  code_review/SKILL.md  — Code review workflow
AGENTS.md               — Project context for the agent