Loading ToolBench
MCP server exposing the research pipeline as a tool. Fast AI research agent in Rust — local GPU or OpenAI
Single tool 'research' has a complete input schema with 5 parameters (query required, mode/domain_profile/domains/intent optional). Description is adequate (95 chars) but generic, lacks WHEN to use, prerequisites, or output structure. Schema is well-formed with proper types and descriptions for each param. However, no output schema is documented, and error handling guidance is absent. The tool does one thing (execute research pipeline) but lacks composition hints for chaining with other tools.
Execute the research pipeline with configurable mode, domain filtering, and intent specification
No output schema documented. LLMs cannot plan downstream tool calls or extract structured data from results.
Tool description lacks WHEN to use it and what it returns. At 95 chars, it's below the 194-char baseline for A+ tools. No guidance on prerequisites (e.g., 'requires SEARXNG_URL env var').
No error handling guidance. If the research pipeline fails (network error, LLM timeout, invalid domain), the tool provides no recovery hints to the LLM.
Parameter 'domains' description says 'Comma-separated domains' but type is array of strings. LLM may pass a single comma-separated string instead of an array, causing parsing errors.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | <=2025-11-25 | v2 |
Parameter 'mode' accepts 'quick, summary, report, deep' but no enum constraint in schema. Free-form string invites hallucinated values like 'fast' or 'detailed'.