What is gcontext
gcontext is a cloud workspace your AI agent connects to over MCP (Model Context Protocol). It gives your agent persistent memory, structure, and the ability to operate your tools, while keeping secrets and execution on your machine.
The core idea
Structure lives in the cloud. Execution stays on your machine.
| In the cloud | On your machine |
|---|---|
| Folders, docs, tasks, integrations | Script execution (Python via uv run) |
| Secret NAMES and descriptions | Secret VALUES (~/.gcontext/.env) |
| Type definitions and templates | Tool output (scrubbed of secrets) |
| The dashboard (visual view of your workspace) | Your AI client (Claude Code, Cursor, etc.) |
Your agent sees the workspace identically from any machine. Your secrets never leave your device.
What you can do with it
- Organize work with typed folders: tasks track progress, integrations connect services, notes hold knowledge your agent can look up later.
- Automate operations with scripts and slash commands: Python that runs locally with secrets injected, promoted from one-off code to reusable commands.
- Build workflows that improve themselves: a support queue where every resolved ticket teaches the agent how to handle the next one.
- See what your agent sees on the dashboard: every folder, every doc, every secret name, rendered as a visual workspace.
How it connects
Your AI client connects to gcontext as an MCP server. The connector (gcontext-mcp on PyPI) runs locally and authenticates with a token from your dashboard.
Your AI client ←→ gcontext-mcp (local) ←→ gcontext cloud (workspace)
↕
your secrets (~/.gcontext/.env)The agent gets tools (tool_read_file, tool_write_file, tool_run_script, etc.), resources (every file is an mcpfs:// resource), and prompts (slash commands, integration loaders).
Ready to connect? Next: Connect your agent.