A tool for toggling MCP (Model Context Protocol) entries in agent configuration files. Supports interactive UI and CLI commands for detecting installed agents, listing MCP entries, and toggling MCP configurations across JSON, TOML, and YAML formats.
mcp-toggler has severe definition quality gaps. Tool names lack action verbs (detect, mcps, toggle, interactive are weak; should be list_agents, list_mcps, toggle_mcp, launch_interactive_ui). Descriptions exist but are generic and lack LLM-optimized structure. Input schemas are completely absent from the source code, no JSON Schema visible for any tool. Parameters are mentioned in descriptions but not formally typed. No output schemas documented. Error handling is minimal. The server reads configuration files and toggles MCP entries, but the tool interface is not production-ready for agent use.
Detect installed agents by checking for configuration files and binaries in standard locations for Claude Code, OpenAI Codex CLI, and Gemini CLI
Launch the interactive terminal UI for browsing and toggling MCP entries with a visual interface
List MCP entries for a specified agent or configuration file. Supports both agent keys (claude_code, codex, gemini_cli) and direct config file paths
Toggle or set a specific MCP entry to enabled or disabled state. Supports forcing enable (--on) or disable (--off) for a given MCP ID in an agent's configuration or a specific config file
No input schemas visible in source code. Tools mcps, toggle, and interactive accept parameters (agent, config, mcp_id, on, off) but no JSON Schema definitions are present. Schema score for all tools is 0.
Tool names lack action verbs. 'detect', 'mcps', 'toggle', 'interactive' are weak. Should be 'list_agents', 'list_mcps', 'toggle_mcp', 'launch_interactive_ui' to signal intent clearly to LLMs.
No output schemas documented. Tools return unstructured text or interactive UI output. LLMs cannot plan downstream calls or extract structured data.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 28 | 2026-07-28+ | v2 |
Descriptions are generic and lack LLM-optimized structure. Missing WHAT/WHEN/WHY guidance. E.g., 'Detect installed agents' (27 chars) does not explain when to call it or what it returns. Baseline is 50-200 chars with clear context.
No error handling guidance. Code calls os.Exit(1) on errors with minimal context. LLMs receive no recovery hints (e.g., 'Agent not found. Try detect() first.').