1. Conduid
  2. Data
  3. io.github.opjhabuilds/designpin-mcp-server
MCP server · Data

io.github.opjhabuilds/designpin-mcp-server

Push HTML prototypes to DesignPin for team review and pull reviewer feedback into your AI.

Unclaimed data
37Low

Scored 5 months ago · breakdown

About io.github.opjhabuilds/designpin-mcp-server

io.github.opjhabuilds/designpin-mcp-server is an MCP server in the Data category: push HTML prototypes to DesignPin for team review and pull reviewer feedback into your AI.

Install

Claude Code
claude mcp add io-github-opjhabuilds-designpin-mcp-server -- npx -y @designpin/mcp-server
npx
npx -y @designpin/mcp-server

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about io.github.opjhabuilds/designpin-mcp-server

Powered by Claude · Grounded in docs

I know everything about io.github.opjhabuilds/designpin-mcp-server. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ✓README presentThe repository ships documentation.

5 checks have not been run yet.

README

DesignPin MCP Server

An MCP server that lets any AI assistant push HTML prototypes to DesignPin for team review and read back reviewer feedback.

What this does

DesignPin is a tool for reviewing HTML prototypes with DOM-anchored comments. This package wraps the DesignPin REST API as a Model Context Protocol server, exposing three tools — createShareLink, uploadVersion, listComments — to any MCP-compatible AI client.

Once configured, your AI assistant — Claude Desktop, Cursor, VS Code (Copilot), Gemini CLI, or any other MCP-compatible client — can ship a generated HTML prototype to a shareable review link, pull back reviewer comments to incorporate, and push iterations as new versions, all from inside the same conversation that produced the design.

For ChatGPT users: ChatGPT does not currently support MCP servers. Use the direct REST API via Custom GPT Actions instead — no MCP server needed.

Install + run

The server is invoked by your MCP client; you don't run it directly during normal use. The simplest path is npx invocation, which fetches and runs without a global install:

npx -y @designpin/mcp-server --api-key dp_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx --project-id proj_example_abc123

For most workflows you'll set this in your MCP client's config file rather than running it manually — see Client config examples below.

To verify the install works:

npx -y @designpin/mcp-server --version
# → 0.1.0

npx -y @designpin/mcp-server --help
# → usage info

Get an API key

  1. Open designpin.pro and sign in.
  2. Open any project (or create one).
  3. Click API & Integrations in the project header.
  4. Click Generate new key, give it a name (e.g. "Claude Desktop").
  5. Copy the key immediately — it's shown once and cannot be retrieved later.

Each key is scoped to a single project. Generate separate keys per client / use case so you can revoke them individually.

Configuration

Both CLI flags and environment variables are supported. Flags win when both are present.

CLI flag Env var Required Default Description
--api-key DESIGNPIN_API_KEY for uploadVersion, listComments — Your dp_live_... API key
--project-id DESIGNPIN_PROJECT_ID for uploadVersion — Project ID the key is scoped to
--base-url DESIGNPIN_BASE_URL no https://designpin.pro API base URL (override for testing)
--help — no — Show usage and exit
--version — no — Print version and exit

Security note: CLI flags are visible in ps output on multi-user systems. Prefer environment variables on shared machines.

Tools

createShareLink

Create a public review link for an HTML prototype. No project setup required — creates a brand-new throwaway project. Rate-limited to 10 requests/hour per IP.

Input Type Required Description
html string yes Complete HTML document to share
title string yes Display name for the share, max 80 chars
authorName string no Author name shown on review page

Returns: JSON with url, reviewToken, projectId, moduleId, versionId.

Example prompt to your AI assistant:

"Take this HTML and create a DesignPin share link titled 'Landing page hero V3'."

uploadVersion

Upload a new HTML version to an existing DesignPin module. Auto-increments versionNumber. Requires the API key + projectId from configuration plus a moduleId from the user.

Input Type Required Description
html string yes Complete HTML for the new version
moduleId string yes Target module within the configured project
description string no Version description shown in the sidebar

Returns: JSON with url, versionId, versionNumber.

Example prompt:

"Push this revised HTML as a new version to module mod_example_def456 with the description 'Address pin #3 contrast feedback'."

listComments

Fetch comments visible on a specific version of a DesignPin module. Uses chronological cutoff semantics matching the web UI: a comment is visible on version V if its origin version is V or any earlier version.

Input Type Required Description
moduleId string yes Module ID
versionId string yes Specific version to view comments on

Returns: JSON with a comments array and a summary string like "3 open, 1 resolved".

Example prompt:

"Get the comments on module mod_example_def456 at version ver_example_ghi789 and tell me what's blocking approval."

Client config examples

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "designpin": {
      "command": "npx",
      "args": ["-y", "@designpin/mcp-server"],
      "env": {
        "DESIGNPIN_API_KEY": "dp_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
        "DESIGNPIN_PROJECT_ID": "proj_example_abc123"
      }
    }
  }
}

Restart Claude Desktop after saving.

Edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "designpin": {
      "command": "npx",
      "args": ["-y", "@designpin/mcp-server"],
      "env": {
        "DESIGNPIN_API_KEY": "dp_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
        "DESIGNPIN_PROJECT_ID": "proj_example_abc123"
      }
    }
  }
}

Reload Cursor after saving.

Add to your VS Code user settings JSON or a workspace .vscode/mcp.json:

{
  "servers": {
    "designpin": {
      "command": "npx",
      "args": ["-y", "@designpin/mcp-server"],
      "env": {
        "DESIGNPIN_API_KEY": "dp_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
        "DESIGNPIN_PROJECT_ID": "proj_example_abc123"
      }
    }
  }
}

Edit your Gemini CLI config (path varies by version). Both the env-style and the flag-style work; flag-style is shown here:

{
  "mcpServers": {
    "designpin": {
      "command": "npx",
      "args": [
        "-y", "@designpin/mcp-server",
        "--api-key", "dp_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
        "--project-id", "proj_example_abc123"
      ]
    }
  }
}

Example conversation

You:  Generate a pricing page with three tiers (Free, Pro $12, Team $29).
      Push it to DesignPin so my team can review.

AI:   [generates HTML, calls createShareLink]
      Share link created. Send this to your team:
      https://designpin.pro/review/EXAMPLE_TOKEN_xyz

      Project: proj_example_abc123
      Module:  mod_example_def456
      Version: ver_example_ghi789

[team reviews and leaves 4 comments]

You:  Pull the feedback from module mod_example_def456 at version
      ver_example_ghi789 and address the must-fix comments.

AI:   [calls listComments]
      Open comments (4):
      - P0 Pin #1 "CTA contrast fails WCAG AA on the Pro tier"
      - P1 Pin #2 "Pricing hidden until hover — make always visible"
      - P1 Pin #3 "Annual toggle is ambiguous — show savings %"
      - P2 Pin #4 "Logo could link back to homepage"

      I'll regenerate the HTML with the two P0/P1 issues addressed.

You:  Push the revision.

AI:   [generates updated HTML, calls uploadVersion]
      v2 uploaded. Same review URL — your team can compare versions.

Direct REST API

If you want to call the DesignPin API without an MCP server, the OpenAPI 3.1 spec is published at:

https://designpin.pro/openapi.json

Import it into Postman, generate a client SDK, or use it directly with ChatGPT Custom GPT Actions.

Links

License

MIT © 2026 Omprakash Jha

README copy captured when this server was indexed; it may lag the repository. The original is authoritative.

Questions

About io.github.opjhabuilds/designpin-mcp-server

How do I install io.github.opjhabuilds/designpin-mcp-server?

Run claude mcp add io-github-opjhabuilds-designpin-mcp-server -- npx -y @designpin/mcp-server, then add the server to your MCP client's configuration.

Is io.github.opjhabuilds/designpin-mcp-server safe to use with an AI agent?

Its trust score is 37 out of 100 (low). It passes 1 of the 1 static security checks run so far. It has no ConduID identity yet, so agent calls to it are not receipted.

Is io.github.opjhabuilds/designpin-mcp-server still maintained?

Conduid hasn't recorded a commit date for this repository yet. Check the repository directly for recent activity.