Loading ToolBench
Multi-service MCP server providing tools for browser automation, Google Workspace, image generation, Korean weather, code execution, document retrieval, and web search
33 tools with highly variable quality. Most have descriptions (good), but schemas are incomplete or missing type information. Many parameters lack descriptions. Tool naming is generally clear (verb_noun pattern), but composition issues exist: browser tools are well-separated, but gog_* tools conflate multiple Google Workspace services. Error handling is minimal, no recovery guidance, no actionable error messages. Output schemas are not documented. Critical issue: parameters like 'account' in gog_* tools lack descriptions; 'strength' in generate_image_from_image is a float with no range guidance. Baseline: 549 production tools average 4 params per tool with 100% param descriptions in A+ tier; this server has ~3.5 params/tool but ~40% lack descriptions.
Go back in browser history.
Click an element by index (from browser_state) or at pixel coordinates.
Hover over an element by index.
Click an element by index and type text into it. Equivalent to clicking the element first, then typing.
Send keyboard keys (e.g. "Enter", "Control+a", "Tab", "Escape").
Navigate to a URL. Starts the browser daemon if not running.
Take a screenshot of the current page.
STDIO transport only, not remotely accessible. Cannot be used by hosted MCP clients or cloud-based agents. Requires local subprocess execution.
Missing parameter descriptions in 13+ tools. Examples: gog_run 'account' param undescribed; generate_image_from_image 'strength' lacks range guidance (0.0 - 1.0); browser_back, browser_type, browser_hover have minimal param docs. LLMs cannot infer parameter meaning from names alone.
No output schemas documented. Tools return results but LLMs cannot plan downstream calls without knowing field names and types. Example: gog_gmail_search returns 'one row per thread' but structure is not specified.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | <=2025-11-25 | v2 |
Scroll the page up or down.
Select a dropdown option by element index.
Get the current page state: URL, title, and clickable elements with their indices. ALWAYS run this first to discover element indices before clicking or typing.
Switch to a browser tab by index.
Type text into the currently focused element.
Generate an image from text using Stable Diffusion 3.5 Large on Amazon Bedrock. SD3.5 Large delivers strong prompt adherence, photorealistic quality, and improved text rendering.
Transform an existing image using Stable Diffusion 3.5 Large (image-to-image). Use this for style transfer, variations, or guided modifications of an existing image.
Show available calendar event colors (IDs 1-11).
List calendar events for a date range.
List Google Contacts.
Read Google Docs content as plain text.
Search Google Drive files.
Search Gmail messages (per email, ignores threading). Use when you need every individual email.
Search Gmail threads by query. Returns one row per thread.
Run a gog CLI command for Google Workspace (Gmail, Calendar, Drive, Contacts, Sheets, Docs).
Get values from a Google Sheet range.
Get Korean air quality forecast from 에어코리아 (airkorea.or.kr). Returns fine dust (PM10, PM2.5) and ozone forecasts.
Get real-time AWS (Automated Weather Station) observations from 기상청 (temperature, wind direction, wind speed, humidity).
Get Korean weather forecast from 기상청 날씨누리 (weather.go.kr). Returns short-term forecast with temperature, precipitation, wind, and humidity.
Get links to Korean weather and air quality forecast pages (기상청 날씨누리, 에어코리아).
Execute Python code to perform calculations or data processing. You MUST provide the 'code' parameter.
Execute a Python script to draw a graph. You MUST provide the 'code' parameter.
Query the keyword using RAG based on the knowledge base.
Performs a web search using Tavily's AI search engine and generates a direct answer to the query, along with supporting search results.
Searches recent news articles using Tavily's specialized news search functionality.
Performs a comprehensive web search using Tavily's AI-powered search engine. Excels at extracting and summarizing relevant content from web pages, making it ideal for research, fact-finding, and gathering detailed information.
No error handling or recovery guidance. Tools fail silently or return raw errors. No actionable messages like 'User not found. Try search_users() first.' Agents cannot self-correct.
Tool composition issue: gog_* tools conflate multiple Google Workspace services (Gmail, Calendar, Drive, Contacts, Sheets, Docs) under one namespace. Consider splitting into separate tool groups (gog_gmail_*, gog_calendar_*, etc.) for clarity and to match the 54-pattern single-responsibility principle.