BabelWrap is a browser automation SDK, not an MCP server. The codebase shows a TypeScript client library (BabelWrap class, Session class) that wraps HTTP calls to a remote API. Tool definitions are inferred from method signatures in src/client.ts and src/session.ts, not explicitly registered as MCP tools. No MCP server implementation, transport layer, or tool registration mechanism is visible. Descriptions are present but generic (10-50 chars, well below the 50-200 char LLM-optimized baseline). Input schemas are partially documented in the rubric but not visible in actual JSON Schema format in the source. Error handling exists (BabelWrapError class) but lacks recovery guidance. No tool annotations, no pagination support, no structured output documentation.

• Build an MCP server wrapper around this SDK. Create a server.ts that implements the MCP protocol, registers these 24 tools with proper JSON Schema definitions, and exposes them via HTTP or SSE transport.
• Expand tool descriptions to 50-200 characters. Include WHAT the tool does, WHEN to use it, and WHAT it returns. E.g., 'Navigate to a URL in the current browser session. Use this to load a new page or follow a link. Returns the page title and URL after navigation completes.'
• Document output schemas for every tool. Define the structure of responses (e.g., listSites returns {sites: [{id, name, url}], total_count}). This enables LLMs to chain tools correctly.
• Add recovery guidance to error messages. Instead of throwing BabelWrapError('invalid_url', 'URL is invalid'), return 'Invalid URL format. Expected http:// or https://. Did you mean: https://example.com?'
• Add pagination support to list tools (listSites, history). Accept limit and offset/cursor parameters. Return total_count and next_cursor. Cap results at 50 items by default to avoid context explosion.