DeepSeek Harness
Run an AI agent on your computer that edits project files and runs commands
by DeepSeek
Codewhale
by CodewhaleAgentWrite code
Lets an AI agent in your terminal edit your code and run commands, with any model
About Codewhale
Codewhale is an open-source coding agent for your terminal: you describe a task and it reads your repository, edits files and runs commands, asking first before shell commands that change things.
For DevelopersTerminal users
Made by Codewhale
Codewhale is published by the codewhale-hq organization on GitHub as an open-source project under the MIT licence.
- Stars
- ★ 41k
- License
- MIT
- Cost
- Usage-based
- Platforms
- Mac · Win · Linux
- Level
- Intermediate
Screenshots
From the project's own repository.
Overview
Codewhale is an open-source coding agent for the terminal. You give it a task; it reads your repository, edits files and runs commands until the task is done or it needs input from you. It works with more than 40 built-in model providers and with models on your own machine, shows file changes as diffs, and keeps a receipts list of what it did, with an undo command. It suits developers who are comfortable in a terminal.
A receipts list shows the files changed, commands run and web calls, and one undo command takes the last request back.
Try these first
Start with the first one: it is a small task you can check easily.
- Explain how to run this project's tests. Don't change any files.First try · read-only
- Fix the failing tests and explain what changed.edits files
- Add a --verbose flag to [command] and run the tests afterwards.edits files
- List the files you changed and the commands you ran in this session.read-only
Example prompts written by Agent Flows Hub.
Features
From the creator's documentation.
| Feature | What it does | Runs |
|---|---|---|
| Task loop | Reads your repository, edits files and runs commands until the task is done or it needs input | Your computer |
| Modes | Plan reads only; Work edits and runs commands; Operate keeps going until the goal is met | Your computer |
| Receipts and undo | Lists files changed, commands run and web or MCP calls; /undo takes back the last request | Your computer |
| Permissions | Ask, Auto-Review or Full Access set when it asks before acting | Your computer |
| Model choice | More than 40 built-in providers, any OpenAI-compatible endpoint, or local models | Your model provider |
| Web search and MCP | Can search the web and call MCP servers you connect | The web |
Good to know
You want an agent that works from your repository in the terminal. Codewhale reads the code, edits files, runs commands and keeps a receipt of what it changed, so you can review the result.
Limits
- In the default Ask permission it applies file edits in your repository without asking and asks only before shell commands that are not read-only, so commit or stash your work first.
- Telemetry is on by default and sends aggregate counts; codewhale config set telemetry false turns it off.
- Windows has no operating-system sandbox yet, and Full Access turns the sandbox off.
Notes by Agent Flows Hub (original project name: Codewhale), written from the creator's documentation, not from running it ourselves. Codewhale README · Codewhale installation guide
Stuck setting it up?
Tell us what happened. We take a look first; if it needs paid work, you get an estimate before anything starts.
More to write code
Other agents
- Codex CLI Lets you hand coding tasks to OpenAI's agent, which edits and runs code on your computer
- Comet Lets an AI assistant in your web browser research, compare pages and handle web tasks
- CowAgent Gives you an AI assistant that plans tasks and replies in WeChat, Feishu, Slack and more
- Cursor Lets you build and change software by describing what you want to an AI agent in an editor
- Devin Takes a coding task from you and works on it until the change is ready to review
- Eigent Hands multi-step office tasks to a team of AI agents in a desktop app on your computer