AI Coder — multi-provider coding assistant CLI (Claude Code compatible). Supports OpenRouter, Anthropic, OpenAI, Gemini, Moonshot Kimi, DeepSeek, Z.AI (GLM), Ollama.
Critical gaps across all dimensions. Tool definitions lack input schemas entirely, no JSON Schema visible for any of the 6 tools. Descriptions are present but minimal (10-60 chars), below the 50-200 char LLM-optimized baseline. No parameter documentation, no output schemas, no error handling guidance. Tools are named after VSCode APIs (VSCodeDiagnostics, VSCodeTasks) rather than action verbs (get_diagnostics, run_task). This is a VSCode extension MCP server, not a general-purpose tool suite, which explains the naming, but the lack of schema and parameter documentation is a hard blocker for LLM integration.
Read the Problems panel diagnostics from the editor's language server
Format a file using the editor's configured formatter
Query the language server for all references to a symbol
Rename a symbol across the codebase using the language server
Run a tasks.json task in the editor
Open a folder as a workspace in the editor
No input schemas visible for any tool. JSON Schema with parameter types and constraints is completely absent.
Tool names do not follow verb_noun convention. Names like 'VSCodeDiagnostics' and 'VSCodeTasks' are noun-based and do not convey action. Should be 'get_diagnostics', 'run_task', 'open_workspace', 'find_references', 'rename_symbol', 'format_file'.
Descriptions are minimal (10-60 chars), well below the 50-200 char LLM-optimized baseline. No explanation of WHEN to use each tool, WHAT it returns, or prerequisites. E.g., 'Read the Problems panel diagnostics from the editor's language server' lacks context on output structure and when to call it.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 28 | 2026-07-28+ | v2 |
No parameter documentation. Tools like VSCodeTasks and VSCodeWorkspace clearly accept parameters (task name, folder path) but no descriptions, types, or constraints are visible. LLMs cannot infer parameter meaning from names alone.
No output schemas documented. LLMs need to know what fields to expect from each tool to plan downstream calls and extract data. E.g., does get_diagnostics return an array of {file, line, message, severity}? Unknown.