Codewhale

// Overview

Codewhale is an open-source coding agent that runs in your terminal. It reads your repository, edits files, runs commands, inspects the results and keeps working toward a goal — either through its TUI or as a one-shot command. It is written in Rust and distributed through npm, Cargo, Docker, Nix, Scoop, prebuilt archives, Android/Termux and a CNB mirror.

The project's distinguishing choice is provider neutrality. You can connect hosted model providers or point it at local models served through Ollama, vLLM or SGLang, and swap provider and model mid-session with the `/model` command. Codewhale began life as `deepseek-tui` and still keeps that configuration and session compatibility, but it is now provider-neutral and independently maintained, and states it is not affiliated with any model provider.

Control over what the agent may do is explicit rather than implicit. Plan mode is read-only; Ask, Auto-Review and Full Access make approval behaviour visible; `/undo` reverts the last turn and `/restore` returns the workspace to an earlier snapshot. Optional OS sandboxing adds a stronger execution boundary where the platform supports it. For longer work, sessions can be saved, a durable `/goal` set, and workflows reviewed before they run. The agent is extensible through MCP servers, skills, hooks and agent roles kept as readable files in project or personal settings.

// What it does

• Provider-neutral: hosted providers or local models through Ollama, vLLM or SGLang, switchable in-session with `/model`
• Graded approval modes — Plan (read-only), Ask, Auto-Review and Full Access — plus optional OS sandboxing
• `/undo` reverts the last turn and `/restore` rolls the workspace back to an earlier snapshot
• Non-interactive `codewhale exec "…"` for scripting a task without opening the TUI
• Extensible via MCP servers, skills, hooks and agent roles stored as readable files
• Saved sessions, a durable `/goal`, and reviewable workflows for long-running work
• Shell completions for bash, zsh, fish, PowerShell and elvish

// Getting started

The quickest route is the npm package; the first run walks you through connecting a provider or staying offline. All commands below come from the project README.

npm install -g codewhale
codewhale

Fix the failing tests and explain what changed.

codewhale exec "fix the failing tests and explain what changed"

codewhale completion zsh

/help
/model

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Reach for it when you want a terminal coding agent that is not tied to one model vendor
• Reach for it when work has to stay local — point it at Ollama, vLLM or SGLang instead of a hosted API
• Reach for it when you need explicit approval boundaries, with an undo and a workspace snapshot to fall back on
• Reach for it to script a one-off repository change from CI with `codewhale exec`

// How Codewhale compares

Codewhale alongside other open-source autonomous coding agents tools AI/TLDR tracks, ranked by GitHub stars.