Automation and Integration
MCP Integration
CloudGrid provides MCP (Model Context Protocol) servers for AI tool integration. This lets AI assistants (Claude, etc.) deploy content to CloudGrid directly.
Two hosted MCP endpoints
| Endpoint | Auth | Use case |
|---|---|---|
mcp.cloudgrid.io | Anonymous-first | Drop files without an account |
mcp-connected.cloudgrid.io | Required (OAuth) | Full access with your identity |
Available MCP tools
| Tool | What it does |
|---|---|
cloudgrid_drop | Upload a file/artifact to the grid (creates an inspiration) |
cloudgrid_claim | Claim anonymous drops after signing up |
cloudgrid_login | Authenticate with CloudGrid |
Connect from Claude Desktop (local edition)
The local MCP edition runs as a stdio subprocess with the full CloudGrid CLI toolset. Configure it in your Claude Desktop settings:
{
"mcpServers": {
"cloudgrid": {
"command": "node",
"args": ["/path/to/mcp-server/src/index.js"]
}
}
}
Connect from claude.ai (web edition)
Add the MCP server URL in your claude.ai settings:
- Anonymous drops:
mcp.cloudgrid.io - Authenticated (full access):
mcp-connected.cloudgrid.io
The connected endpoint triggers OAuth when authentication is needed.
Usage examples
Once connected, you can interact with CloudGrid through natural language:
“Drop this HTML page onto CloudGrid, visible to anyone with the link.”
“Publish this React component to my team’s grid.”
“Upload this analysis as a PDF to CloudGrid.”