MCP Server: Overview & Setup

The Stringhive MCP server lets AI coding assistants talk to your translation data directly.

Use cases

• Ask your AI assistant "what locales are missing translations in the my-app hive?" and get an answer
• Push new source strings after adding a feature, without leaving your editor
• Pull translated files and check completion before a release

Connecting

The MCP endpoint is at:

https://www.stringhive.com/mcp

Authentication uses the same API tokens as the REST API. Create a token in Settings > API Tokens with the abilities you need (read for browsing, write for pushing strings or translations).

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "stringhive": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://www.stringhive.com/mcp"],
      "env": {
        "STRINGHIVE_TOKEN": "shv_yourtokenhere"
      }
    }
  }
}

Restart Claude Desktop and you should see the Stringhive tools available.

Cursor

In your .cursor/mcp.json:

{
  "mcpServers": {
    "stringhive": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://www.stringhive.com/mcp"],
      "env": {
        "STRINGHIVE_TOKEN": "shv_yourtokenhere"
      }
    }
  }
}

Other MCP clients

Any client that supports the MCP protocol can connect. Pass your token via the STRINGHIVE_TOKEN environment variable or directly in the transport config, depending on what the client supports.

Permissions

The MCP server respects the same token abilities and Hive restrictions as the REST API. A read-only token can browse and export but can't push strings. A token scoped to specific Hives can't see others.