ChromaDB Remote MCP provides 19 tools with generally complete schemas and descriptions. All tools have input schemas with typed parameters and descriptions, placing this above the median for community servers. However, several quality gaps prevent a higher score: (1) output schemas are not documented anywhere in the visible source, (2) tool descriptions vary in quality and some lack actionable context, (3) error handling guidance is minimal, (4) no tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk stratification, (5) parameter descriptions could be more specific about formats and constraints. The server correctly identifies risk levels (READ_ONLY, WRITE, DESTRUCTIVE) but does not expose this via tool annotations. Composition is clean, tools have single responsibilities and names follow verb_noun convention. Parameter naming is mostly consistent (collection_name appears across tools, supporting tool chaining). For a vector database server, the tool set is comprehensive, covering CRUD operations, collection management, and admin/distributed variants.

• Document output schemas for all tools. For example, chroma_list_collections should define: { collections: [{ id: string, name: string, metadata?: object }], total_count: integer, next_cursor?: string }. Include all fields returned in the response.
• Add tool annotations to the schema. Mark destructive tools (chroma_delete_collection, chroma_delete_documents, chroma_admin_delete_*) with destructiveHint=true and idempotentHint=false. Mark read-only tools with readOnlyHint=true. This enables clients to enforce safety policies.
• Specify enum values and constraints in parameter descriptions. E.g., for 'include' parameter: 'Array of response fields to include. Valid values: documents, metadatas, distances, embeddings. At least one must be specified.'
• Add min/max bounds to numeric parameters. E.g., 'limit: Maximum number of results to return (1 - 1000, default 20)'. This prevents LLMs from passing unreasonable values.
• Document error conditions and recovery paths for each tool. E.g., 'If both query_embeddings and query_texts are provided, query_texts is ignored. If collection does not exist, returns error: {code: "COLLECTION_NOT_FOUND", message: "Collection {name} does not exist. Call chroma_list_collections to discover available collections."}'
• Clarify mutual exclusivity in parameters. For chroma_add_documents, explain whether documents array can be empty if embeddings are pre-computed, or whether at least one document is always required.
• Add dependency hints for discovery tools. E.g., chroma_list_collections description: 'Call this first to discover available collections before querying or modifying them.'

Inferred effective spec: 2026-07-28+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry