Single tool (get_server_info) with clear naming and good description (156 chars). Input schema is empty object (valid for no-param tool). Description explains purpose, when to use, and explicitly states token is never included. However, output schema is not documented in visible code, only inferred from _build_server_info() implementation. Tool is read-only and self-describing, but lacks formal output schema documentation that LLMs need to plan downstream actions. No error handling guidance visible.

• Document the output schema for get_server_info in the tool registration. Include field names, types, and descriptions: {read_only: boolean, read_tool_count: integer, mutating_tool_count: integer, server_url_source: enum('token'|'override'), hatchet_sdk_version: string, python_version: string}.
• Add error handling guidance to get_server_info description: 'If the tool returns unexpected values, the Hatchet SDK may not be initialized or the server may be in an inconsistent state. Check server logs and retry.'
• Migrate from STDIO to Streamable HTTP transport to enable remote access and compatibility with hosted MCP clients.
• For each of the 56 tools, ensure descriptions are 50-200 characters, all parameters have type + description, and output schemas are documented in the tool registration (not just inferred from implementation).
• Verify all mutating tools carry destructiveHint=True and idempotentHint appropriately set. Document which tools support dry-run or confirmation patterns.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry