Loading ToolBench
MCP server for Parsnip memory management platform - a knowledge graph engine for managing entities, observations, relationships, and tags across projects
Parsnip MCP provides 13 well-structured tools with complete JSON schemas and clear descriptions. All tools have input schemas with typed parameters and descriptions. Tool names follow verb_noun convention (search_knowledge, create_entities, delete_relations). Descriptions are concise (50-150 chars) and explain purpose. However, output schemas are not documented, LLMs cannot predict response structure. Parameter descriptions lack actionable constraints (e.g., fuzzyThreshold 0.0-1.0 is stated but no guidance on when to use fuzzy vs exact). Error handling is minimal, no recovery guidance or categorization. Some tools (delete_entities, delete_observations) lack confirmation/dry-run patterns for destructive operations. Batch operations are well-designed (create_entities, add_observations accept arrays), reducing multi-call overhead.
Add new observations to existing entities.
Add categorical tags to existing entities for filtering and organization.
Create new entities with observations and optional tags. Use a single call for multiple entities.
Create directional relationships between entities. Supports cross-project relations by specifying fromProjectId/toProjectId.
Permanently delete entities and all their relationships.
Delete specific observations from entities while preserving the entity.
Delete specific relationships between entities.
Output schemas not documented. LLMs cannot predict response structure for any tool. Responses for read_graph, open_nodes, traverse_graph are opaque, agents cannot plan downstream calls or extract required fields.
Destructive operations (delete_entities, delete_observations, delete_relations) lack confirmation or dry-run patterns. Agents can permanently delete data without a safety gate.
No error handling guidance. Tools do not document what errors are retryable, user-fixable, or fatal. LLMs receive no recovery hints (e.g., 'Entity not found. Try search_knowledge() first.').
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | <=2025-11-25 | v2 |
List all available projects.
Retrieve specific entities by exact names along with their relationships.
Retrieve the complete knowledge graph for a project.
Remove tags from entities.
Search entities by text or tags across projects. Omit projectId to search all projects.
Traverse the knowledge graph starting from specified entities with configurable depth and direction.
Parameter constraints lack actionable guidance. fuzzyThreshold (0.0-1.0) and depth (1-10) have ranges but no explanation of when to use each value or what happens at boundaries.
Tool descriptions do not state whether operations modify state. Agents cannot distinguish safe (read-only) from unsafe (write/delete) tools without reading the Risk field, which is metadata, not description.