Codewhale

Codewhale is an open-source coding agent for your terminal, written in Rust. It works with any model, can split one task across several agents, and every line of it is yours to read.

curl -fsSL https://codewhale.net/install.sh | sh

This installs the latest release on macOS or Linux, and it needs no account. Windows, npm and Cargo are below.

Website · Documentation · Changelog · Contributing

简体中文 · 日本語 · Tiếng Việt · Bahasa Indonesia · 한국어 · Español · Português · Русский · Українська · Français · Deutsch · 繁體中文 · हिन्दी · Türkçe · Italiano · Polski · العربية · Català

A development build of 0.10.3, which is not released yet, with a scripted test model, in Full Access inside a throwaway workspace. The edit, three tests and a sub-agent review ran for real.

Install

The command above puts checksum-verified codewhale and its short alias codew in ~/.local/bin. If your shell then says command not found, follow the PATH note the installer prints (details). codewhale update upgrades it.

winget install --id HunterBown.CodeWhale --exact --source winget  # Windows x64
npm install -g codewhale                                          # wraps the same release binaries
cargo install codewhale-cli --locked                              # builds from source; can trail GitHub Releases

Prebuilt binaries cover Linux, macOS and Windows on x64 and arm64. Scoop, Nix, Homebrew on Linux, Android and Termux (preview) and manual downloads are in the installation guide, and containers are in Docker. Use one install method per machine; otherwise two copies end up on your PATH and whichever comes first wins.

Quickstart

1. Commit or stash your work, then run codewhale from the root of a git repository. By default it applies file edits there without asking and asks before each shell command that is not read-only.
2. Run /provider (or press F3), pick a provider, paste its API key and choose a model. DeepSeek is the default provider. On a first launch with no key, Codewhale asks for a provider when you send your first message, and it offers a local model when a local Ollama server answers.
3. Type a task:

Fix the failing tests and explain what changed.

Read-only commands run straight away. For any other shell command Codewhale asks first: y runs it once, a allows commands like it for the rest of the conversation, and n refuses. /undo takes back your last request and its file changes, and /help lists commands and keys.

For scripts and CI, codewhale exec runs one task without the interface. Read headless runs before you give it tools with --auto, because --auto uses Full Access unless you add a sandbox.

What it does

Give Codewhale a task and it reads the repository, edits files and runs commands until the task is done or it needs your input. The latest release notes list the newest changes, and the changelog has every version.

Any model

More than 40 providers are built in, including Anthropic, OpenAI, DeepSeek, Google, Mistral, Moonshot, OpenRouter and xAI. Any OpenAI-compatible endpoint works, and so do local models through Ollama, vLLM or SGLang. The default model is DeepSeek's deepseek-flash, which now serves DeepSeek V4.1 Flash (details).

/model switches models between turns, and /model auto picks one for each turn (/router sets up how). /turn inspect shows which model a turn used, what it cost and how long it took. See Providers.

Receipts and undo

/receipts, or codewhale receipts --last after you quit, lists the files a session changed, the commands it ran, its web and MCP calls, the agents it started, each approval and who gave it, and what failed. /undo takes back your last request and the files it changed in one step. /restore rolls the whole workspace back to an earlier snapshot, and because that overwrites later edits it needs /trust on or Full Access. See Receipts.

Several agents on one task

/goal sets an objective that carries across turns. Codewhale can split work across sub-agents, run a Fleet of agents with their own roles and models, or follow a workflow checked into the repository.

Extensions

Connect MCP servers, add skills and plugins, and run hooks on session and tool events. Code mode (experimental, on by default) lets the model write one execute_tools program that calls several MCP and plugin tools; each call still meets your approval rules and keeps its receipt. Computer Use is a separate plugin released on its own (experimental, macOS only), and stays off until you review its access and turn it on.

This is the beginning of the README. Read the rest at the original. Read the full README ↗