Loading ToolBench
MCP Server for Personal Brain API - provides semantic search and archival capabilities for documents and chat sessions through Claude Desktop
This server has moderate definition quality with significant gaps. All 9 tools have descriptions and most have parameter schemas, but many descriptions are generic and lack the specificity needed for LLM decision-making. Parameter schemas are present but incomplete, several parameters lack type information or constraints. No tool annotations (readOnlyHint/destructiveHint) despite clear semantic differences (e.g., upload_document and save_chat are write operations). Output schemas are not documented for any tool. Error handling guidance is absent. The naming is generally clear (verb_noun pattern), but descriptions fall short of the 50-200 char LLM-optimized range and lack dependency hints or prerequisites.
Ask a question and get an answer with proper citations from your knowledge base.
Get detailed information about a specific document including all chunks.
List all uploaded documents with their metadata.
Retrieve saved chat conversations from your knowledge base.
Save a chat conversation to your personal knowledge base.
Search through archived chat sessions.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing what fields are returned. This is a critical gap affecting tool chainability.
No tool annotations despite clear semantics. write-operation tools (save_chat, upload_document) lack destructiveHint=true. Read-only tools (search_*, list_*, get_*) lack readOnlyHint=true. LLMs cannot determine tool safety without explicit hints.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 56 | <=2025-11-25 | v2 |
| 2026-03-09 | D | 54 | - | v1 |
Search through uploaded documents with semantic search.
Search across all your documents and conversations using semantic search.
Upload and process a document for semantic search.
Descriptions are too brief and lack context. Baseline for A+ tools is 50-200 chars; most descriptions here are 30-60 chars, leaving LLMs unable to determine when to select each tool vs similar alternatives (e.g., search_documents vs search_memory vs search_chat_history).
Parameter descriptions lack constraint information. E.g., top_k range (1-20) is in description but format is inconsistent. Content_type, tool, and tags parameters lack enum values or valid examples. LLMs will hallucinate invalid values without explicit constraints.
No error handling guidance. If search_documents returns no results, list_all_documents fails, or ask_with_citations hits an API timeout, the LLM has no recovery path. Errors must include 'try this next' suggestions.
No distinction between search_documents and search_memory tools. Both accept a query parameter and return results, but the descriptions do not explain when to use each (documents-only vs combined documents+chats). LLMs will waste reasoning cycles deciding.
Tool responses likely lack IDs needed for chaining. E.g., if search_documents returns results but omits document_id, the LLM cannot call get_document_details. If ask_with_citations returns citations but omits document IDs, followup calls are blocked.