Rail Connector MCP has 13 tools with complete input schemas and descriptions. Naming follows verb_noun convention (start_remote_control, send_text, get_session_transcript). Descriptions are present and actionable (avg ~80 chars), explaining WHAT each tool does. However, parameter descriptions are minimal, most lack context about format, constraints, or dependencies. Output schemas are not documented in the source. Error handling is basic (errorReporting=true but no recovery guidance visible). No tool annotations (readOnlyHint/destructiveHint). The server targets a specialized domain (Claude Code remote control) with clear intent, but lacks production-grade polish in parameter validation and error messaging.

• Document output schemas for all tools. E.g., status() returns {broker_status: string, managed_sessions: [{name: string, state: string}], capabilities: {...}}. This enables LLMs to chain tools and extract required fields.
• Expand parameter descriptions with format/constraint details. E.g., 'permissionMode (enum: default|manual|acceptEdits|auto|bypassPermissions|dontAsk|plan), use "manual" to require user approval before Claude makes edits; use "auto" for hands-off mode.'
• Add tool annotations: mark start_remote_control, stop_remote_control, send_text, send_key, submit_prompt as destructiveHint=true; mark status, get_claude_capabilities, list_sessions, capture_terminal as readOnlyHint=true.
• Include recovery hints in error scenarios. E.g., 'If managedSession not found, call list_sessions(archiveState="active") to see available sessions, then retry with the correct name.'
• Define managedSession parameter format explicitly: 'A session identifier (string, typically "rail-connector-managed" or a custom name). Use list_sessions() to discover available sessions.'