OpenFGA MCP server has 10 tools with complete input schemas and descriptions in Korean. All tools are properly registered with enum-constrained environment parameters. However, descriptions lack depth and actionability, most are 20-60 chars, below the 50-200 char baseline for LLM optimization. No output schemas are documented, forcing LLMs to guess response structure. Parameter descriptions are minimal (e.g., 'DSL 파일 경로' without format guidance). Error handling is absent from tool definitions. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels. Batch tool (openfga_tuple_batch_write) is present but single-item variants dominate, missing composition optimization.

• Add output schemas to all tools. Document response structure: field names, types, and whether results are paginated. Example: openfga_store_list should return {stores: [{id, name, created_at}], total_count, next_cursor}.
• Expand descriptions to 80-150 chars. Include WHAT the tool does, WHEN to use it, and key output fields. Example: 'List all OpenFGA authorization stores in the specified environment. Returns store IDs and metadata. Use this to discover available stores before reading models or tuples.'
• Add tool annotations: mark openfga_tuple_delete and openfga_model_write with destructiveHint=true; mark read-only tools (openfga_check, openfga_tuple_read, openfga_store_list) with readOnlyHint=true.
• Document parameter constraints in descriptions. For filePath: 'File path to .fga DSL file (UTF-8 encoded, max 1MB)'. For dsl: 'OpenFGA DSL string (see https://openfga.dev/docs/dsl-syntax for syntax rules)'.
• Add error handling guidance to tool descriptions. Example for openfga_model_write: 'On syntax error, returns the line number and error message. On permission error, returns 403 with required scopes. Retryable on 5xx errors.'