On this page
One agent. A workspace for everyone.
Claude Code is Anthropic’s coding tool for reading a project, changing files and running commands. Crux provides the editor, shared text, spatial canvas and terminal around your work.
In a desktop build with terminal support, a command-line agent runs as a separate process. Install and authenticate it using the provider’s instructions. Opening a project in Crux does not create a Claude account or grant access to a model.
Start with one change you can review.
A task such as “add jitter to retry backoff and test the failure cases” has a useful boundary. The agent can work through a small set of files, and a teammate can understand the resulting diff.
- Open the repository and inspect its instructions and test commands.
- Name the behavior you want and the files or function it relates to.
- Let the agent inspect the implementation before proposing changes.
- Review the diff, run the focused tests, then check the wider project where needed.
Keep the explanation next to the change.
Use the canvas to arrange related files when a change crosses a boundary: a request handler, its helper and a test, for example. Discuss those connections with collaborators instead of relying on a list of filenames alone.
Shared text helps collaborators see changes. It does not establish that a generated change is correct. Read the diff and check the actual command output before treating the task as complete.
Bring your own configured workflow.
Crux is in private alpha. Join the waitlist to request access, and check the available runtimes in your build. Claude Code installation, authentication and permissions follow Anthropic’s documentation.
BUILD SOMETHING TOGETHER
Your team.
Your agents.
One workspace.
Bring your people, code and context together in Crux.
Join the waitlist