Laver MCP has 13 tools with consistent naming (verb_noun pattern: workspace_read, board_read, ticket_create, etc.) and complete input schemas with typed parameters. Descriptions are present but brief (10-80 chars), below the 50-200 char LLM-optimized baseline. Parameter descriptions are minimal or absent in several tools. Output schemas are not documented. Error handling exists but lacks recovery guidance. No tool annotations (readOnlyHint, destructiveHint) despite clear risk stratification in the tool list.

• Expand tool descriptions to 50-200 chars. Example: 'workspace_read' → 'Retrieve the workspace name, UUID, and list of all boards. Call this first to discover available boards before reading or creating tickets.'
• Document output schemas for all tools. Example: 'board_read returns {board_uuid, name, description, columns: [{name, tickets: []}]}'. Include field types and which fields are needed for downstream tool calls.
• Add tool annotations: mark ticket_delete, wiki_delete, attachment_delete with destructiveHint=true; mark all read tools with readOnlyHint=true. This enables agent safety checks.
• Enhance error messages with recovery guidance. Example: 'Ticket not found (404). Try board_read(board_uuid) to list available tickets, or ticket_create() to make a new one.'
• Clarify attachment_upload: specify file parameter format (base64-encoded string? raw bytes?), max size, and accepted MIME types. Add description: 'Upload a file (max 50MB, any MIME type) to attach to a ticket or wiki page.'