Loading ToolBench
95 typed Typesense operations for AI agents through an MCP server that is read-only by default.
TypesenseKit MCP exposes 75 tools with minimal descriptions (most 10-30 chars, well below the 194-char baseline). Tool names follow verb_noun convention (list, retrieve, create, update, delete, search) which is good, but descriptions are generic placeholders lacking WHEN/WHY context. No visible input schemas in the provided source code, cannot verify parameter types, constraints, or descriptions. Output schemas are undocumented. Error handling is absent from tool definitions. The server implements tool annotations (readOnlyHint, destructiveHint, idempotentHint) which is a positive signal, but without schemas and rich descriptions, agents cannot reliably select or invoke tools. This is a classic case of API-to-MCP wrapper that exposes raw operations without LLM-optimized definitions.
Create an alias
Delete an alias
List all aliases
Retrieve an alias
Update an alias
List analytics events
Make a raw API call to Typesense
Create a new collection
Delete a collection
No input schemas visible in source code. Cannot verify parameter types, constraints, or descriptions for any of 75 tools. Per HARD SCORING RULE, schema score must be 0 for tools without visible schemas.
Descriptions are generic placeholders (10-30 chars). Baseline is 194 chars. Examples: 'List all collections', 'Retrieve a collection schema', 'Search documents in a collection'. These lack WHEN to use, prerequisites, and expected output structure. Per HARD SCORING RULE, descriptions under 20 chars score 0-20.
No output schemas documented. Agents cannot plan downstream tool calls or extract required fields (e.g., does search return result count? pagination cursor? field names?). Violates pattern:tool and pattern:paginated-result.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 32 | 2026-07-28+ | v2 |
List all collections
Retrieve a collection schema
Update a collection schema
Wait for collection schema changes
Create a conversation
Delete a conversation
List all conversations
Retrieve a conversation
Update a conversation
Create a curation set
Delete a curation set
List all curation sets
Retrieve a curation set
Update a curation set
Delete documents from a collection
Export documents from a collection
Import documents into a collection
Retrieve a single document
Search documents in a collection
Upsert documents into a collection
Create an API key
Delete an API key
List all API keys
Retrieve an API key
Update an API key
List all natural language search models
Retrieve a natural language search model
Configure slow request tracking
List slow requests
Create an override
Delete an override
List all overrides
Retrieve an override
Update an override
Create a preset
Delete a preset
List all presets
Retrieve a preset
Update a preset
Execute multiple searches
Create a snapshot
List snapshots
Create a stemming dictionary
Delete a stemming dictionary
List all stemming dictionaries
Retrieve a stemming dictionary
Update a stemming dictionary
Create a stopword set
Delete a stopword set
List all stopword sets
Retrieve a stopword set
Update a stopword set
Create a synonym set
Delete a synonym set
List all synonym sets
Retrieve a synonym set
Update a synonym set
Create a synonym
Delete a synonym
List all synonyms
Retrieve a synonym
Update a synonym
Get debug information
Check cluster health
Get cluster metrics
Get cluster statistics
api.call tool is dangerously generic ('Make a raw API call to Typesense'). Marked IRREVERSIBLE risk but no confirmation/dry-run pattern. Violates pattern:confirmation-request for destructive operations.
No error handling guidance in tool definitions. Agents cannot distinguish retryable errors from user-fixable ones. Violates pattern:recovery-guide and pattern:error-classification.