codewhale-hq/Codewhale

What it solves

Codewhale is an open-source AI agent designed to automate software development tasks. It solves the problem of manually editing files, running terminal commands, and verifying code changes by allowing an AI to interact directly with your project repository and local environment.

How it works

The tool operates as a runtime that connects to a model of your choice (hosted or local via Ollama, vLLM, or SGLang). Users can interact with it through a terminal interface (TUI), a local web client, or a VS Code extension. The agent can read the codebase, edit files, and execute shell commands to achieve a specific goal. It supports different permission modes (Ask, Auto-Review, and Full Access) to ensure the user maintains control over the agent's actions.

Who it’s for

It is built for developers who want an AI-powered coding assistant that can operate autonomously within their local environment while remaining model-agnostic and providing granular control over execution permissions.

Highlights

• Model Agnostic: Connect to various hosted providers or local runtimes like Ollama.
• Autonomous Execution: Capable of reading repositories, editing files, and running commands.
• Extensible: Supports MCP servers, custom skills, and hooks.
• Computer Use: Includes a plugin for observing and interacting with other applications on the OS.
• Flexible Interfaces: Available via TUI, web client, and VS Code extension.