LetCode is an interactive console AI agent designed for software development directly within your terminal. It supports a wide range of popular LLM providers and allows the AI to interact with your project's file system through integrated tools for reading, writing, and executing commands
- Multi-Model Support: Native support for many providers, including OpenRouter, Gemini, Anthropic, DeepSeek, OpenAI, Mistral, Qwen, Hugging Face, Groq, NVIDIA, MiniMax, Xiaomi, and xAI
- Autonomous Agent: The AI can read files, create/edit code, and execute shell commands within your project context
- Interactive Interface: User-friendly terminal interface with menu navigation and real-time visualization of the AI's "thought" process
- Security: Configurable confirmation prompts before performing critical operations (file read/write, command execution)
- Flexibility: Multilingual interface support, configurable reasoning effort, and seamless switching between models
The following commands are available during a session:
| Command | Description |
|---|---|
/key |
Interactive API key setup for the active provider |
/model <name> |
Change and save the model |
/provider |
Interactive provider selection |
/effort <val> |
Configure reasoning effort (low, medium, high) |
/language |
Switch interface language |
/reset |
Reset current chat history |
/settings |
Configure confirmation settings for tools |
/help |
Show list of available commands |
/exit |
Exit the program |
- Configuration: Settings and API keys are stored in json file
- Security: The
execute_command,read_file, andwrite_filetools are scoped to the working directory (defaults to~/Desktopor the current directory) - Synchronization: The agent supports interactive command execution with user-interrupt capabilities
This project is licensed under the MIT License - see the LICENSE file for details.