Loading ToolBench
Personal AI assistant — synthesizes activity across multi-project folders, Brave bookmarks, Telegram saved messages, and a MemPalace memory layer. Provides read/write palace access for Claude Code with semantic search, memory management, and synthesis tools.
MemPalace MCP exhibits significant definition quality gaps. Tool names follow verb_noun convention (mempalace_status, mempalace_search, mempalace_add_drawer) which is good, but descriptions are extremely terse (10-50 chars), well below the 194-char baseline. Most critically, 5 of 12 tools have NO input schema visible in the source code (whats_new, surface_ideas, take_in_work, mempalace_reconnect, mempalace_status, mempalace_list_wings, mempalace_get_taxonomy). Parameters that ARE defined lack depth, no enums for constrained values, no format specifications, no dependency documentation. Error handling is absent from visible code. Output schemas are undocumented. The synthesis layer tools (whats_new, surface_ideas, take_in_work) are imported but their actual tool registration and schemas are not visible in the provided source, forcing a cap at 50 per the inference rule.
file verbatim content into a wing/room
check if content already exists before filing
remove a drawer by ID
full wing → room → count tree
rooms within a wing
all wings with drawer counts
force cache invalidation and reconnect after external writes
semantic search, optional wing/room filter
total drawers, wing/room breakdown
5 tools have no visible input schema (whats_new, surface_ideas, take_in_work, mempalace_status, mempalace_list_wings, mempalace_get_taxonomy).
All tool descriptions are under 50 characters (baseline 194 chars). Examples: 'total drawers, wing/room breakdown' (34 chars), 'all wings with drawer counts' (28 chars).
No output schemas documented. LLMs cannot infer what fields to expect from tool responses (e.g., does mempalace_search return drawer_id, content, wing, room, score?). This violates the 100% baseline for A+ tools.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | <=2025-11-25 | v2 |
surface ideas from synthesis layer across multi-project folders, Brave bookmarks, and Telegram exports
take in work from synthesis layer across multi-project folders, Brave bookmarks, and Telegram exports
synthesize what's new across multi-project folders, Brave bookmarks, and Telegram saved messages
No error handling guidance visible. Tools like mempalace_delete_drawer (destructive) and mempalace_add_drawer (write) lack recovery instructions. No distinction between retryable vs fatal errors.
Synthesis layer tools (whats_new, surface_ideas, take_in_work) are imported but their actual tool registration, schemas, and parameter definitions are not visible in the provided source. Per inference rule, these cap at 50.