Most people think of a GitHub repository as a place to store code. Claude Spaces are something different: they're pre-configured workspaces where the repository itself is the application. There's no build step, no deployment pipeline. You clone the repo, open Claude Code, and you have a working AI workspace tuned for a specific purpose.
A Claude Space typically includes a CLAUDE.md file for instructions, slash commands for common operations, MCP server configurations for external integrations, subagent definitions for specialised tasks, and structured data directories that Claude reads from and writes to. The key insight is that Claude Code isn't just a coding assistant — it's a general-purpose agent that happens to live in your terminal.
The Claude Agent Workspace Model is the reference template that all these workspaces follow. It defines the folder structure, Claude Code scaffolding, and conventions for turning a repo into a purpose-built AI workspace for any domain — not just code. Want to create your own? Point Claude at this repo and ask it to scaffold one for your use case.
View Model on GitHub → Learn more →
Use Claude Spaces as system administration interfaces — describe what you want in natural language and Claude handles the execution across your infrastructure.
Explore →Forkable templates for common workflows — daily planning, budgeting, communications, writing, home automation, and document analysis.
Explore →The reference template and documentation for building your own agent workspaces. Point Claude at the model repo to scaffold new workspaces.
Explore →The complete specification: required folder structure, the four-layer stack, lifecycle stages, context architecture, and validation checklist.
Read the spec →When to use a portable plugin versus a full agent workspace — and the rationale for migrating between them.
Learn more →