MCP Server

Model Context Protocol is Anthropic's open standard for connecting AI assistants to data sources and tools. orbitnest_mcp is the official MCP server for OrbitNest — install it once and your AI agents can manage projects, query data, and deploy functions safely.

What is MCP?

MCP is a standard protocol AI tools use to talk to external systems. Instead of your AI guessing at shell commands or HTTP calls, it discovers a curated set of tools with typed inputs and outputs — and invokes them on your behalf.

With the OrbitNest MCP server installed, Claude Desktop (or Cursor, or any MCP-capable client) knows how to:

• Sign in as an admin and list projects
• Create, inspect, and delete tables
• Run SQL queries and inspect results
• Enable/disable RLS, create and manage policies
• Deploy edge functions and manage environment variables
• Upload files, list buckets, manage storage
• Rotate API keys, configure JWT settings, read logs

Install

Claude Desktop

Add the OrbitNest MCP server to your Claude Desktop config file:

// ~/Library/Application Support/Claude/claude_desktop_config.json
// (on Windows: %APPDATA%/Claude/claude_desktop_config.json)
{
  "mcpServers": {
    "orbitnest": {
      "command": "npx",
      "args": ["-y", "@orbitneststudio/mcp-server"],
      "env": {
        "ORBITNEST_API_URL":    "https://studio.orbitnest.io",
        "ORBITNEST_ADMIN_EMAIL": "you@example.com"
      }
    }
  }
}

Restart Claude Desktop. The OrbitNest tools appear in the hammer icon. Sign in once with orbitnest_admin_signin (email + password) — your token is cached for the session. If you signed up with Google, GitHub, or Apple and have no password, use orbitnest_login instead: it opens your browser to complete the OAuth flow and captures the session automatically.

Cursor

Add the same config to Cursor's MCP settings (Command Palette → “MCP: Edit Configuration”).

A typical AI session

Once installed, you can simply tell the AI what you want. It picks the right MCP tools automatically.

You:  Create a "posts" table with id, title, body, and author_id
      referencing auth.users. Enable RLS with a policy that lets
      authors update their own posts.

AI:   [calls orbitnest_execute_sql with the CREATE TABLE statement]
      [calls orbitnest_enable_rls on posts]
      [calls orbitnest_create_policy with the USING clause]

      Done. I created the posts table, enabled RLS, and added a policy
      "authors manage own posts" scoped to the owner via auth.uid().

Available tools

The MCP server exposes 60+ tools grouped by module:

• Auth — signin, login (Google/GitHub/Apple OAuth), signup, signup_verified, signout, refresh, profile
• Projects — list, get, create, update, delete, set_active
• Database — execute_sql, create_table, get_schema, get_table_data, insert_row, update_row, delete_row, bulk_insert/update/delete
• RLS — enable_rls, disable_rls, list_policies, create_policy, delete_policy
• Functions — list_functions, create_function, update_function, delete_function, get_function_logs
• Env vars — get_env_variables, create/update/delete_env_variable
• Storage — list_buckets, create_bucket, upload_file, download_file, list_files, delete_files
• API Keys — get_api_keys, create_api_keys
• Logs — query_logs, get_log_stats, get_slow_queries, export_logs
• Admin — list_admins, create_admin, update_admin, delete_admin

Safety & scoping

Best practices:

• Use per-project admin accounts with minimal permissions
• Keep debug mode enabled for easy OTP flows during testing
• Review the Logs view after AI-driven changes
• Use a staging project for destructive explorations

Open source

The orbitnest_mcp source is available on GitHub. Fork, extend, or use it as a reference for building your own MCP servers.