Loading ToolBench
MongoDB MCP server providing secure access to MongoDB databases
Three tools with complete schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools are read-only (safe), but descriptions lack actionable context for LLM selection. Parameter descriptions are present but generic. No output schemas documented. Error handling exists but lacks recovery guidance. Naming is clear and verb-based (aggregate, explain, sample).
Run a MongoDB aggregation pipeline
Get the execution plan for an aggregation pipeline
Get random sample documents from a collection
Output schemas not documented. LLMs cannot predict what fields to expect from aggregate(), explain(), or sample() results, forcing them to guess at downstream field names and risk broken chains.
Pipeline parameter description is generic ('MongoDB aggregation pipeline stages (e.g., $match, $group, $sort)'). Does not explain format constraints, valid stage types, or error handling for invalid stages. LLMs may pass malformed pipelines.
No error recovery guidance. Error responses (e.g., invalid collection name, pipeline syntax error) do not tell the LLM what to do next or how to self-correct. Raw MongoDB errors leak to the agent.
Sample tool description lacks context for when to use it vs. aggregate(). Does not explain use case (exploratory analysis, schema discovery) or when to call it in a workflow.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Options parameter in aggregate() is under-specified. allowDiskUse, maxTimeMS, and comment are documented, but no guidance on when to use them, performance implications, or defaults. LLMs may omit critical options.