Loading ToolBench

Reading the current corpus…

design-mcp MCP server · F · ToolBench

design-mcp

MCP server for AI-assisted design generation, refinement, and canvas management. Generates HTML designs from prompts using Claude, manages design sessions, and provides design artifact storage.

2026-07-28+STREAMABLE_HTTP11 tools

45/100

Definition Quality59

Protocol Readiness48

Supportability21

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence · evidence: stateless requests, Streamable HTTP

Current-spec patterns detected

Summary

Server has 11 tools with reasonable naming (verb_noun pattern) and descriptions present for all. However, parameter schemas are incomplete or missing for several tools. Input schemas visible in tool definitions show basic structure but lack comprehensive type annotations and constraints. Output schemas are not documented. Error handling is minimal, no recovery guidance or actionable error messages visible in source. Tool descriptions average ~120 chars, which is acceptable but could be more specific about when/why to use each tool. No tool annotations (readOnlyHint/destructiveHint) despite clear READ_ONLY vs WRITE distinctions.

Tools (11)

create_editor_sessionwriteauth50/100

Create a new interactive editor session for collaborative design editing.

extract_canvas_designread onlyauth50/100

Extract design tokens and structure from an existing canvas or HTML document.

generate_designwriteauthsource verified77/100

Generate a complete HTML design from a prompt, with optional repository context, viewport specification, and aesthetic direction control.

get_canvasread onlyauth50/100

Retrieve a specific canvas (saved design) by ID.

get_credit_statusread onlyauth50/100

Retrieve the current credit balance and usage statistics for the authenticated user.

link_editor_sessionwriteauth50/100

Link an editor session to a design run, enabling live editing and synchronization.

list_canvasesread onlyauth50/100

List all design canvases (saved designs) for the authenticated user.

Issues (5)

HIGHtool ↗

Output schemas not documented. Tools return JSON payloads (e.g., generate_design returns run_id, html, summary) but no formal schema is visible in tool definitions. LLMs cannot plan downstream calls without knowing response structure.

generate_designrefine_designlist_canvasesget_canvasextract_canvas_design

HIGHtool-annotations ↗

No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Risk field indicates READ_ONLY vs WRITE, but MCP tool annotations are absent. Agents cannot infer safety/idempotency from tool definitions alone.

generate_designrefine_designcreate_editor_sessionlink_editor_sessionunlink_editor_sessionrate_style_direction

HIGHrecovery-guide ↗

Minimal error handling guidance. No visible recovery hints, error categorization, or actionable messages. If a tool fails, LLM has no guidance on retry strategy or next steps.

generate_designrefine_design

Recommendations

Add formal output schemas to all tools. Document return types as JSON Schema (e.g., generate_design returns {run_id: string, html: string, summary: string, cost_usd: number}).
Annotate tools with readOnlyHint/destructiveHint. Mark get_*, list_*, whoami as readOnlyHint=true; mark generate_design, refine_design, create_*, link_*, rate_* as destructiveHint=true.
Expand parameter descriptions with constraints. E.g., 'limit (1-100, default 20)', 'feedback (max 2000 chars)', 'rating (1-5 scale)'.
Add recovery guidance to error responses. E.g., 'Design generation timed out. Retry with a shorter prompt or simpler viewport. If persistent, check credit balance with get_credit_status.'
Enhance tool descriptions with context. E.g., 'whoami: Returns authenticated user profile (name, email, tier). Call this first to verify credentials before expensive operations like generate_design.'

Score history

Overall score trend

First recorded score · v2 rubric

45/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	45	2026-07-28+	v2