Loading ToolBench
6-layer intelligent web intelligence platform for scraping, analyzing, synthesizing, and visual auditing of web content with local and cloud LLM integration
24 tools with mixed quality. Naming is verb-forward and mostly clear (scrape, analyze, synthesize, screenshot, vision_analyze, get_*, update_*, list_*, delete_*, add_*, search_*). Descriptions present for all tools (range 50 - 250 chars, baseline 194). However, parameter descriptions are sparse or missing entirely. Input schemas visible but incomplete: many parameters lack type definitions (e.g., 'array of strings' instead of {"type": "array", "items": {"type": "string"}}). Output schemas not documented. Error handling guidance absent. Security concerns: update_config accepts API keys as parameters (pattern:secret-injection violation). No pagination guidance for list_history. Composition is reasonable (tools are single-purpose), but chaining IDs not always returned.
Register custom LLM API provider
Register custom OpenRouter model
Analyze raw text with LLM using smart routing between Ollama (local) and OpenRouter (cloud)
Remove custom API provider
Remove custom OpenRouter model registration
Retrieve current server configuration with masked API keys
Retrieve scraping history with optional filtering
API keys exposed as tool parameters in update_config. Credentials must use server-side secret injection via environment variables, not tool parameters. Agent traces log all parameters, secrets leak into logs and prompt history.
Input schemas incomplete: parameters use informal type descriptions ('array of strings', 'integer') instead of proper JSON Schema with type, items, and format fields. Prevents LLM validation and schema-aware clients from working correctly.
Output schemas not documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., does scrape return extracted_text, summary, entities?). Breaks tool chaining and forces agents to guess response structure.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Discover site structure and generate sitemap
Get aggregate statistics on scraping activity
Get service status including Ollama, OpenRouter, Scrapling, and custom API availability
List available vision models (Ollama + OpenRouter)
List all available models (Ollama local + OpenRouter + custom APIs)
Delete Ollama model
Pull Ollama model with streaming progress (Server-Sent Events)
Single URL scraping with optional LLM analysis. Supports multiple strategies (free, fast, smart, llm, ai, premium, stealth) and tasks (summarize, extract_entities, classify, etc.)
Parallel scraping of multiple URLs with concurrent execution control
Capture screenshot of a URL with optional PDF export and element waiting
Full visual audit of a website across multiple pages
Capture screenshots of multiple URLs with optional vision analysis
Search OpenRouter model catalog
Generate website code/spec from multiple URLs with LLM synthesis
Generate multiple variants of website code from URLs
Update server configuration at runtime
Analyze screenshot with vision LLM (local Ollama or cloud OpenRouter)
No error handling guidance. Tools lack recovery hints (e.g., 'If URL is invalid, check format' or 'If Ollama unavailable, ensure ollama serve is running'). Agents cannot self-correct on failures.
Destructive tools (ollama_delete, delete_openrouter_model, delete_custom_api) lack confirmation or dry-run support. Agents can permanently delete models without safeguards.