MCP server for controlling Tabby terminal via Chrome DevTools Protocol
Server has 4 tools with complete input schemas and descriptions. Naming follows verb_noun convention (list_targets, execute_js, query, screenshot). Descriptions are present and adequate (80-180 chars), exceeding the 20-char minimum. All parameters have type definitions and descriptions. However, output schemas are not documented, LLMs cannot predict response structure. Error handling is minimal: no recovery guidance, no categorization of retryable vs fatal errors. The 'target' parameter accepts both int and string (index or WebSocket URL), which is good for natural identifiers, but the description could be clearer about the resolution logic. No tool annotations (readOnlyHint, destructiveHint) despite clear risk profiles (execute_js is WRITE, others are READ_ONLY). Missing pagination guidance for list_targets. Overall solid foundation but lacks production-grade error handling and output documentation.
Execute JavaScript code in Tabby terminal context and return the result. Code is wrapped in async IIFE by default for fresh scope and await support.
List available CDP targets (tabs) with their index, URL, and WebSocket URL
Query DOM elements by CSS selector. Automatically waits for Angular and element to exist.
Capture screenshot of Tabby terminal window or specific element
Output schemas not documented. LLMs cannot predict response structure or plan downstream tool calls. list_targets returns JSON array of targets; execute_js returns TextContent; query returns DOM element data; screenshot returns ImageContent or file path. All should be explicitly documented in tool descriptions or a separate schema section.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). execute_js is marked WRITE risk but has no destructiveHint annotation. list_targets, query, screenshot are READ_ONLY but lack readOnlyHint. Annotations help LLMs reason about safety and retry logic.
Error handling lacks recovery guidance. Code validates inputs (target, code, format, quality) but returns generic ValueError messages. No guidance on what to do next: 'Invalid target: must be int or WebSocket URL from list_targets()' would be actionable; current 'target must be int or str' is not.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | B | 70 | <=2025-11-25 | v2 |
list_targets has no pagination or limit guidance. If Tabby has many tabs, response could be large. Description should state 'Returns all available targets' and recommend pagination if needed.
Target parameter description could clarify resolution logic. Currently: 'Target tab: index (0=first, -1=last) or WebSocket URL from list_targets'. Should add: 'If passing index, 0 is first tab, -1 is last. If passing WebSocket URL, must match output from list_targets().'