What is CoCalc?
CoCalc is a platform where users describe a software task and an AI agent builds it within a project containing their files and installed tools. It supports the integrated Codex agent, Claude Code, or other command-line agents running in project terminals. Claude Code availability depends on site configuration and requires a CoCalc membership, a Claude Pro or Max subscription, or an Anthropic API key.
What are the features of CoCalc?
- AI code generationUsers describe a desired outcome and point the agent at project files; the agent works using the software already installed in that project.
- Side-by-side reviewGenerated applications, reports, or charts open beside the conversation so users can inspect results and ask the agent to explain its actions.
- Iterative refinementUsers can request alternative versions from the agent or open files, notebooks, and terminals to continue editing manually.
- Multiple agent supportThe platform runs an integrated Codex agent, Claude Code where enabled, or other command-line agents directly in project terminals.
- Live collaborationProject collaborators see edits made by the agent or other users in real time.
- Version restorationUsers can restore earlier versions of their project files.
What are the use cases of CoCalc?
- Application buildingA user describes an application idea and the agent generates the working app within the existing project environment.
- Data analysisA user directs the agent to project files to produce a report or chart based on the available data.
How to use CoCalc?
- Describe the outcomeState what you want and point the agent at your project files via the web interface.
- Review the resultOpen the generated app, report, or chart beside the conversation and ask the agent what it did.
- Refine or take overRequest another version from the agent, or open the files, notebooks, and terminals to continue work yourself.









