Single tool 'extract' has a clear verb-noun name and adequate description (98 chars), but critical gaps undermine quality. Input schema is present with three parameters (base_id, output_dir, cookie), all with descriptions. However, the schema lacks type declarations visible in the MCP registration, only the raw Python function signature is evident. No output schema is documented. The tool accepts a raw cookie as a parameter, violating secret-injection patterns. No error handling guidance, no pagination, no structured output schema. The description does not explain what the tool returns or when to use it. This is a single-tool server with foundational issues.

• Move cookie authentication to server-side secret injection. Accept base_id and output_dir as parameters; inject cookie from environment variable or MCP server config. Return structured error if cookie is missing or invalid.
• Document output schema: return a JSON object with fields like {success: bool, automations_count: int, files_written: [string], errors: [{automation_id, reason}]}. Include this in tool description.
• Add explicit JSON Schema to tool registration with type declarations: base_id (string, pattern: ^app[A-Za-z0-9]{14}$), output_dir (string, minLength: 1), and remove cookie parameter entirely.
• Enhance error handling: return structured errors with recovery hints. E.g., {error: 'Unauthorized', code: 401, hint: 'Cookie expired. Get a fresh one from DevTools → Network tab → copy Cookie header.'}
• Expand tool description to 100-150 chars explaining: 'Extracts all automations (workflows) from an Airtable base by fetching metadata and deployment configs via the internal v0.3 API. Writes JSON files to output_dir. Requires a valid authenticated session cookie.'