VULK MCP server has 9 tools with complete input schemas and descriptions. Naming follows verb_noun convention (generate, edit, list, get, files, deploy, models, usage, subscribe). Descriptions are detailed (100-300 chars) and explain WHAT the tool does and WHEN to use it. However, output schemas are NOT documented, responses are inferred from code comments rather than explicit schema declarations. Parameter descriptions are present but lack format constraints (e.g., projectId format, prompt length limits). Error handling exists but lacks recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in metadata.

• Add explicit output schema definitions for all 9 tools. Document response structure (e.g., generate returns {projectId, editorUrl, previewUrl, filesGenerated, files[], tokens, cost}). Use JSON Schema or TypeScript interfaces visible in code.
• Implement tool annotations: mark generate/edit/deploy with destructiveHint=true; mark list/get/files/models/usage/subscribe with readOnlyHint=true. This enables LLM safety reasoning and prevents accidental misuse.
• Convert 'model' parameter to an enum: ['claude-sonnet-4-20250514', 'gpt-4o', 'gemini-2.5-pro', 'deepseek-chat']. Add prompt length constraint (e.g., 'min 10 chars, max 5000 chars') to parameter description.
• Enhance error messages with recovery guidance. Instead of 'Generation failed: <error>', return 'Generation failed: <error>. Retry with a simpler prompt or check your API key at https://vulk.dev/settings/api-keys'.
• Add pagination to list and files tools: accept optional limit (default 20, max 100) and offset parameters; return total count and next_offset in response. Document in tool descriptions.