Loading ToolBench
A Cursor-inspired AI assistant that offers smart autocomplete and interactive chat with selected notes, with MCP server integration for tool use and resource access
This MCP server has severe definition quality issues across nearly all tools. Most tool definitions lack proper input schemas, descriptions are vague or missing, and parameter documentation is sparse. The server mixes MCP protocol tools with Composio API wrappers without clear separation. Many tools appear to be inferred from code snippets rather than explicitly registered with complete schemas. The architecture suggests this is an Obsidian plugin that bridges to MCP servers rather than a proper MCP server itself.
The duckduckgosearch class utilizes the composio duckduckgo search api to perform searches, focusing on web information and details. it leverages the duckduckgo search engine via the composio duckduckgo search api to retrieve relevant web data based on the provided query.
The eventsearch class enables scraping of google events search queries. it conducts an event search using the composio events search api, retrieving information on events such as concerts, festivals, and other activities based on the provided query.
Perform a search to find similar links and retrieve a list of relevant results. the search can optionally return contents.
The financesearch class utilizes the composio finance search api to conduct financial searches, focusing on financial data and stock information. it leverages the google finance search engine via the composio finance search api to retrieve pertinent financial details based on the provided query.
The googlemapssearch class performs a location-specific search using the composio goolge maps search api. this class extends the functionality of the base action class to specifically target locations related to the given query. by utilizing the google maps search engine through the composio goolge maps search api, it fetches the most relevant location data based on the input query. the `googlemapssearch` class is particularly useful for applications that need to retrieve and display location information about a specific area. it leverages the powerful search capabilities of google's maps search engine, ensuring that the returned results are accurate and relevant.
14 of 25 tools (56%) have NO visible input schema definition. Tools like access_mcp_resource, read_file, write_to_file, list_files, search_files, dataview_query, execute_command, insights, ask_followup_question, attempt_completion, switch_mode, manage_files lack schema documentation.
Description quality is poor across 11 Composio search tools (DUCK_DUCK_GO_SEARCH, EVENT_SEARCH, EXA_SIMILARLINK, FINANCE_SEARCH, GOOGLE_MAPS_SEARCH, IMAGE_SEARCH, NEWS_SEARCH, SCHOLAR_SEARCH, SEARCH). Descriptions are generic, lack information on WHEN to use each tool vs alternatives, and many fail to explain the distinction between similar tools (all are search variants). The generic descriptions do not guide LLM tool selection when multiple search options exist.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 45 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 20 | 1.12.1+ | v1 |
The imagesearch class performs an image search using the composio image search api, to target image data and information. it uses the google images search engine through the composio image search api to fetch relevant image information based on the input query.
The newssearch class performs a news-specific search using the composio news search api. this class extends the functionality of the base action class to specifically target news articles related to the given query. by utilizing the google news search engine through the composio news search api, it fetches the most relevant news articles based on the input query. the `newssearch` class is particularly useful for applications that need to retrieve and display the latest news articles about a specific topic. it leverages the powerful search capabilities of google's news search engine, ensuring that the returned results are current and relevant.
Scholar api allows you to scrape results from a google scholar search query. the scholarsearch class performs an academic search using the composio scholar search api, academic papers and scholarly articles. it uses the google scholar search engine through the serp api to fetch relevant academic information based on the input query.
Perform a google search using the composio google search api.
Convert a document to markdown using the Marker API. Supports PDF, Word, Excel, PowerPoint, HTML, and EPUB files.
Convert a video url, like youtube url to markdown. Supports youtube, bilibili, tiktok, douyin, kuaishou, etc.
Access resources provided by connected MCP servers to retrieve data from various sources.
Generate and apply unified diff patches to markdown files with support for intelligent matching and sub-hunk strategies.
Ask clarifying questions during task execution.
Complete and finalize tasks.
Query data using Obsidian Dataview plugin.
Execute shell commands.
Generate insights and analysis.
List files in a directory.
Manage files including creation, deletion, and organization.
Read files from the filesystem.
Search for files matching patterns.
Switch between different operational modes.
Request to use a tool provided by a connected MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.
Write content to files.
Tool naming violates verb_noun convention. COMPOSIO_SEARCH_* tools use a prefix+API pattern (COMPOSIO_SEARCH_DUCK_DUCK_GO_SEARCH) that is overly verbose and obscures the actual verb. LLMs struggle to infer action intent from names like 'COMPOSIO_SEARCH_SCHOLAR_SEARCH' vs a clearer 'search_scholar' or 'search_academic_papers'. Additionally, generic 'SEARCH' tool name lacks specificity about what is being searched.
No error handling documentation visible across all tools. Tools provide no guidance on what errors might occur, whether they are retryable, or what the LLM should do next. For example, CONVERT_DOCUMENT does not explain what happens if the file format is unsupported or the file is corrupt. apply_diff does not document merge conflict behavior.
EXECUTE_COMMAND is a destructive tool that accepts arbitrary shell commands but lacks input validation, rate limiting, and permission checks. No mention of sandboxing, timeout limits, or what permissions the agent has. This is a critical security risk, agents can cause severe damage.
Missing output schema documentation for almost all tools. Tools do not declare what fields they return, making it impossible for LLMs to plan downstream calls. For example, CONVERT_DOCUMENT presumably returns markdown text, but the structure is undocumented. search_* tools do not document pagination fields, result count limits, or the structure of returned search results.
Tool definitions are partially inferred from code locations rather than from explicit tool registration. Many tools (e.g., read_file, write_to_file, list_files) are referenced in tool-groups.ts but no explicit schema or registration object is visible in the provided source.
Parameter descriptions are sparse. For example, CONVERT_DOCUMENT's 'file_type' parameter description says 'File type extension (pdf, docx, xlsx, etc.) without the dot' but does not specify if the list is exhaustive or if custom types are supported. COMPOSIO_SEARCH_EXA_SIMILARLINK has 11 parameters with varying description quality; several parameter descriptions contain example values (e.g., '2023-01-01T00:00:00Z') which LLMs tend to reuse literally rather than adapt.
No parameter constraints (enums, ranges, pattern regexes) visible for most tools. For example, CONVERT_DOCUMENT accepts file_type as a free-form string, no enum constraint. COMPOSIO_SEARCH_EXA_SIMILARLINK's 'type' parameter accepts 'keyword', 'neural', or 'magic' but is not declared as an enum in the schema (would need additionalProperties constraint in schema to be visible). search tools accept arbitrary query strings with no validation.
No pagination or result-limiting guidance documented for search tools. The Composio search APIs likely return variable result counts. Tools do not advertise pagination parameters, result limits, or how to handle large result sets. This risks context window exhaustion when search results are large.