MCP server for Ultrathink — save and search your knowledge base from Claude Desktop, Claude Code, and any MCP-compatible client.
Two tools with complete schemas and descriptions, but significant gaps in parameter documentation and error guidance. Tool names follow verb_noun convention (ultrathink_save, ultrathink_search). Descriptions are adequate (100-150 chars) but lack WHEN/WHY context. Input schemas are present with types and enums, but parameter descriptions are minimal (10-30 chars). No output schema documented. Error handling returns raw error messages without recovery guidance or categorization. Missing idempotency hints for the write operation.
Save an entry to Ultrathink knowledge base. Use this to store notes, ideas, links, snippets, and other knowledge.
Search the Ultrathink knowledge base using natural language. Returns an AI-generated answer based on stored entries.
Output schemas not documented. LLMs cannot plan downstream operations or extract required fields. ultrathink_save returns {success, message, id, title, type} and ultrathink_search returns {success, answer, matchingTimestamps, entriesSearched}, but these are not declared in tool definitions.
Parameter descriptions are too brief (10-30 chars). 'Title of the entry', 'Notes or description', 'Natural language search query' lack context on format, constraints, or when to use each field. LLMs cannot infer whether 'notes' vs 'content' should be used.
ultrathink_save lacks idempotency guidance and confirmation pattern. This is a destructive write operation (saves to knowledge base), but no dry-run, confirmation, or idempotency hint is offered. Agents retrying on ambiguous failures risk duplicate entries.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
Error handling returns raw error messages ('Error: <error.message>') without recovery guidance. No categorization of retryable vs user-fixable vs fatal errors. LLMs receive no actionable next steps when API calls fail.
ultrathink_save 'type' enum includes 11 values (note, link, idea, snippet, para, outlook, chatgpt, claude, perplexity, notion, long-note, profile) but descriptions do not explain when to use each. LLMs will guess. 'url' is marked optional but description says 'required for link type', undocumented dependency.