Build apps and automate workflows

Codewhale is an open-source agent that writes code, runs commands and works with the tools you connect. Use your existing model APIs, or run models locally and on your own servers.

Explore the terminal

Captured views from a Codewhale demo session. Explore the conversation, task workbar, model connections and help.

What you can build and automate

See every provider

Build apps and useful tools

Create a new application, extend a project, turn data into a report, or write a script that saves you time. The agent can create files, run commands and test the result.

Work across your apps

Add tools through plugins and MCP servers, or use APIs from your own scripts. Each service needs its own setup and authentication.

Make a workflow you can reuse

Run tasks from scripts or CI with codewhale exec. For larger jobs, save a Fleet of agents with different roles and models and bring the team back when you need it.

Work across files and tools

Work with your project, connected services, browser tools and agent teams from one session. Choose the model, add the connections you need, and return to the work without starting over.

See all integrations
Files and terminal
Create files, run commands, inspect data and test what you build. You set the working folder and permissions.
Plugins and connected apps
Add skills and tools through plugins and MCP. Review and enable the connections you want the agent to use.
Browser and computer · preview
Use browser tools and the Computer Use plugin for work in apps and websites, with the access you grant.
Sessions you can return to
Keep the conversation, tool results and work history together. Resume the task in the terminal or its local browser client.
Teams of agents
Use Fleet to split a larger job between agents with different roles and models, and follow their progress in one place.

Set permissions and approvals

Use Plan to explore, Work to carry out a task, and Operate to coordinate larger jobs. Approval settings decide which actions wait for you.

Plan
Blocks file mutation and shell execution. Permitted research may contact external services; session state can still be saved.
Work
Edits files and runs commands within the permission you set.
Operate
Coordinates larger jobs, delegating to other agents when useful. Uses the same tools and permission boundaries as Work.
Ask
Prompts according to the active approval rules; saved permissions and hard policy boundaries still apply.
Auto-Review
Automatically reviews eligible actions and reports any action it cannot approve.
Full Access
Reduces approval prompts. It does not bypass hard policy boundaries or grant access outside the allowed scope.

Available now and in development

The terminal, local browser client, and community CodeWhale GUI are available now and need no Codewhale account. The desktop and hosted web apps are in development, share the same session model, and use an account; your model connection stays your choice.

Terminal
ReleasedInstall release binaries for Linux, macOS, or Windows; npm and Cargo also work, and Android on Termux is a preview. The interactive TUI and codewhale exec for scripts ship together. Install guide
Local browser
Included with the terminalRun codewhale web to open Codewhale in your browser. Read the conversation, send a task, and answer approvals on your machine. Local browser guide
CodeWhale GUI (VS Code)
AvailableA separate community project: chat, threads, and file changes in a VS Code sidebar over the local Runtime. Install it from the VS Code Marketplace; the source is on GitHub. VS Code Marketplace
Hosted web app
Development previewBeing rebuilt to match the desktop app. Today you can sign in, then type /rc in a running terminal session to continue it on the web; hosted task execution is still being qualified. Sign in
Desktop
Development buildThe native app becoming the main Codewhale client: folders, conversations, and model connections in one window. No public download yet.
Cloud computers
In developmentHosted computers that run your tasks.