A version of command line interface(CLI) that leverages the power of Claude Opus(3-4.7). This framework lets Claude create and manage its own tools on the fly, continuously growing its capabilities through the course of a conversation
SaferSkills independently audited claude-engineer (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
An interactive self imporving AI coding agent for creating and using AI tools with Claude.At its core, Claude Engineer introduces a dynamic tool ecosystem where the AI can autonomously create, manage, and execute its own tools during conversations. This allows the system to continuously expand its capabilities, adapting to user needs in real time and becoming more effective the more it is used.
requirements.txtgit clone https://github.com/bawadou/claude-engineer
cd claude-engineer
python -m venv venv
venv\Scripts\activate.bat
pip install -r requirements.txt
# or
pip install -e .
python main.pygit clone https://github.com/bawadou/claude-engineer
cd claude-engineer
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# or
pip install -e
python3 main.pyCreate .env file and copy .env.example to .env and fill in your key:
ANTHROPIC_API_KEY=sk-ant-...The agent supports configuration options through the Config class:
| Option | Description |
|---|---|
CE_MODEL | Which Claude model to use for the conversation |
ANTHROPIC_API_KEY | Your Anthropic API key from console.anthropic.com. The agent will refuse to start without it. |
CE_MAX_TOKENS | Maximum number of tokens the model is allowed to generate in a single response. Higher values allow longer outputs but cost more. |
CE_VERBOSE | When set to 1, prints every tool invocation and its arguments as they happen. Useful for debugging the agent loop. |
CE_TOOLS_DIR | Directory for tool storage |
python -m claude_engineerType /help inside the REPL to see commands.
There are some pre-built tools available in Claude Engineer.
| Tool | What it does |
|---|---|
read_file | Read a text file from the project |
create_tool | Create new tools using frameworks self-improvement core |
write_file | Create or overwrite a text file |
list_dir | List a directory |
run_shell | Run a shell command (60s timeout) |
web_search | Search the web via Anthropic's hosted tool |
├── claude_engineer/ # Main package
│ ├── __init__.py # Package exports
│ ├── __main__.py # Entry point for `python -m claude_engineer`
│ ├── agent.py # Core conversation loop and tool orchestration
│ ├── cli.py # Interactive REPL and argument parsing
│ ├── config.py # Configuration loading and defaults
│ └── tools/ # Built-in tool implementations
│ ├── __init__.py # Tool registry
│ ├── base.py # Abstract tool interface
│ ├── fs.py # File system operations
│ ├── shell.py # Shell command execution
│ └── search.py # Web search integration
├── tests/ # Unit tests
├── pyproject.toml # Package metadata and dependencies
├── requirements.txt # Runtime dependencies
├── .env.example # Template for environment variables
├── .gitignore
└── README.mdContributions are welcome. Whether it's a new tool, or a bug fix — open an issue first if the change is non-trivial, so we can align on the approach before you submit a Pull-Request.
MIT - Use freely, modify as needed, contribute back if you can.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.