Run MCP servers in The Cloud

The Cloud (thecloud.so) is an AI-native workspace and operating system: your memory, pages, apps, and agents live in one always-on place that any AI can read and act in over MCP. That makes it a natural place to run MCP — the server side and the client side live in the same workspace.

MCP runs both ways here

Most setups treat MCP as a plugin you bolt onto one assistant. The Cloud treats it as infrastructure. It is an MCP server: any AI connects and reads your memory, pages, tasks, and calendar, and acts in your workspace with your consent. It is also an MCP client: your workspace and agents reach outside services — GitHub, Gmail, Google Calendar — from the same place.

The result is one always-on workspace in the middle, instead of a mesh of per-assistant plugins.

The MCP server: one endpoint

The Cloud exposes one HTTP endpoint, speaking JSON-RPC 2.0. This is the only address an AI needs:

https://mcp.thecloud.so/mcp

Auth is a browser sign-in over OAuth. Your browser opens once, you sign in to your Cloud account, and the connection is live. No API key is copied, pasted, or stored. For CI and headless environments, a static API key generated in the app works as a Bearer header instead.

What a connected AI can reach

Once connected, an AI works with your workspace through grouped tools: Pages, Memory, Tasks, My Day, Calendar, Email, GitHub, plus Habits, Files, and Profiles. Read tools return your data. Write tools create or change it and hand back a link.

Every tool carries a safety annotation — reads, ordinary writes, and destructive actions are each marked, so a client knows what is safe to run and what to confirm first. What an AI creates defaults to private. Making anything public always requires your explicit confirmation in the app; an AI never publishes on its own.

The MCP client: your workspace reaches out

The same workspace consumes outside services. With GitHub connected, your AI can list repositories, issues, and pull requests, search code, read files, create branches, commit to feature branches, and open pull requests — merging always stays with you. Gmail and Google Calendar connect the same way, so email and scheduling are reachable from the same session.

Agents dispatched to machines you attach use these connections too, so a run can go from your prompt to an open pull request without a local toolchain.

Start here

Claude Code — one terminal command:

claude mcp add --transport http the-cloud https://mcp.thecloud.so/mcp

Cursor — paste into ~/.cursor/mcp.json(global) or a project's .cursor/mcp.json:

{
  "mcpServers": {
    "the-cloud": {
      "url": "https://mcp.thecloud.so/mcp"
    }
  }
}

Claude Desktop — Settings, Connectors, Add custom connector, then paste the endpoint and sign in when prompted.

The full connection guide — Gemini CLI, Windsurf, ChatGPT, API keys, and troubleshooting — lives at thecloud.so/connect.