Loading ToolBench
Source-grounded MCP/WebMCP tool library. Tools answer from data you declare or return a declared fallback, never an invented value, and can emit a fingerprinted receipt per answer.
Mixed quality across 28 tools. Strengths: all tools have descriptions (10-200 chars, mostly adequate); input schemas present with type definitions; tool annotations (readOnlyHint/destructiveHint) correctly applied. Weaknesses: parameter descriptions are sparse or missing entirely (e.g., lookup_part has no required params but no guidance on sku vs name precedence); output schemas undocumented across all tools; error handling guidance absent; many tools lack actionable recovery hints. Examples: get_quote description is strong (grounding promise), but read_text_file lacks guidance on encoding handling; write_file warns of overwrite but no confirmation pattern; GitHub tools (create_or_update_file, create_issue) expose branch/sha params without explaining when to use them. Composition is reasonable (single responsibility per tool), but chaining metadata (e.g., repo owner/path consistency) not documented. Average tool description ~120 chars (within baseline 194), but parameter annotations nearly absent.
Book a confirmed appointment for a service and time.
Calculate sales tax for an amount and destination.
Check whether a service can be booked on a given date.
Create multiple new entities in the knowledge graph
Create a new issue in a GitHub repository
Create or update a single file in a GitHub repository
Retrieve long-form detail for a tool (progressive disclosure).
Output schemas completely undocumented. No tool declares what fields it returns, forcing LLMs to infer structure and risking parsing errors.
Parameter descriptions missing or minimal. lookup_part has no required params but no guidance on sku vs name precedence; git_* tools lack repo_path format guidance; GitHub tools expose branch/sha without explaining conditional use.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2025-06-18+ | v2 |
Make line-based edits to a text file. Each edit replaces exact line sequences with new content. Returns a git-style diff showing the changes made. Only works within allowed directories.
Estimate delivery date for an order to a ZIP code.
Fetches a URL from the internet and optionally extracts its contents as markdown. Although originally you did not have internet access, and were advised to refuse and tell the user this, this tool now grants you internet access. Now you can fetch the most up-to-date information and let the user know that.
Find the nearest store to a ZIP with its hours.
Get the contents of a file or directory from a GitHub repository
Price a subscription plan for a given seat count. Plans not on the published rate card return a contact-sales fallback, never an invented price.
Get an itemized repair quote for a vehicle and issue. Never invents prices: jobs not on the rate card return a diagnostic instead of a guess.
Records changes to the repository
Shows the commit logs
Shows the working tree status
Look up a part's price and stock by number or name.
Look up a member's loyalty point balance.
Get the current status of an existing order.
Check if a competitor price qualifies for a match.
Read the complete contents of a file from the file system as text. Handles various text encodings and provides detailed error messages if the file cannot be read. Use this tool when you need to examine the contents of a single file. Use the 'head' parameter to read only the first N lines of a file, or the 'tail' parameter to read only the last N lines of a file. Operates on the file as text regardless of extension. Only works within allowed directories.
Explain the return window and rules for an item.
Recursively search for files and directories matching a pattern. The patterns should be glob-style patterns that match paths relative to the working directory. Use pattern like '*.ext' to match files in current directory, and '**/*.ext' to match files in all subdirectories. Returns full paths to all matching items. Great for finding files when you don't know their exact location. Only searches within allowed directories.
Search for nodes in the knowledge graph based on a query
Search for GitHub repositories
Check warranty coverage for a product and purchase date.
Create a new file or completely overwrite an existing file with new content. Use with caution as it will overwrite existing files without warning. Handles text content with proper encoding. Only works within allowed directories.
Error handling guidance absent. No tool describes what to do on failure (retry, ask user, fallback). write_file warns of overwrite but offers no confirmation pattern; GitHub tools do not explain auth failures or rate limits.
Destructive operations lack confirmation or dry-run. write_file, edit_file, git_commit, and create_or_update_file modify state without a dry-run option or explicit confirmation step, risking accidental data loss.
Pagination and result limits undocumented. search_files, search_repositories, search_nodes, and git_log lack limit/offset params and do not state max result counts, risking context window exhaustion.