Subgraph MCP demonstrates moderate quality with consistent tool naming and present-but-basic descriptions. All 9 tools follow verb_noun naming conventions (get_*, execute_*, search_), which is strong. However, descriptions are terse (13-87 chars, well below the 194-char baseline), lacking context on WHEN to use each tool or WHAT distinguishes variants. Parameter schemas are present and properly typed (string, object, array), but descriptions are minimal or generic ('The deployment ID (e.g., 0x...)' duplicates the parameter name). No output schemas are documented, agents cannot plan downstream tool composition. Error handling exists but is generic (McpError::invalid_params, McpError::internal_error) without recovery guidance. The server is read-only (all tools marked READ_ONLY), which simplifies safety but limits utility. STDIO-only transport will cap protocol readiness at 50.

• Expand tool descriptions to 100 - 200 chars explaining WHEN to use each tool. E.g., 'Get the GraphQL schema for a specific subgraph deployment. Use this when you have the exact deployment ID (0x...) and want to see all available fields and types. Use get_schema_by_subgraph_id if you only have the subgraph name.'
• Add actionable parameter descriptions: 'deployment_id (string): The on-chain 0x-prefixed identifier for a specific subgraph deployment version. Different from subgraph_id (resolves to latest) and ipfs_hash (content hash). Found via get_top_subgraph_deployments or search_subgraphs_by_keyword.'
• Document output schemas for each tool. E.g., get_schema_by_deployment_id returns { schema: string (SDL), deployment_id: string, network: string }. Include any pagination fields if applicable.
• Consider consolidating the three get_schema_* tools into one with a 'lookup_type' enum (deployment_id|subgraph_id|ipfs_hash) and 'lookup_value' string. Reduces cognitive load and tool explosion.
• Enhance error responses with recovery hints: 'Deployment 0x1234... not found. Try search_subgraphs_by_keyword to find available deployments, or use get_top_subgraph_deployments with a contract address.'
• If possible, migrate from STDIO to Streamable HTTP for remote accessibility and hosted client support.
• Document which tools return paginated results and how to iterate (offset, limit, total_count, or cursor). E.g., 'get_top_subgraph_deployments returns the first 20 results; pass offset=20 for the next page.'

Inferred effective spec: <=2025-11-25.

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