Pix3lMCP

Pix3lMCP is an MCP (Model Context Protocol) server that exposes Pix3lBoard as tools for Claude Code and Claude.ai. It allows your AI assistant to create, read, update, and delete boards, lists, and cards in Pix3lBoard — without leaving your workflow.

Repository: github.com/Pix3ltools-lab/pix3lmcp

Tech Stack

Language: TypeScript
Protocol: Model Context Protocol (MCP)
Transport: Streamable HTTP (stateless)
Vercel variant: Next.js 15 API route
Docker variant: Express server
Validation: Zod
Build: esbuild (Docker), Next.js build (Vercel)

Available Tools

Pix3lMCP exposes 17 tools:

Tool	Description
`list_workspaces`	List accessible workspaces
`list_boards`	List boards in a workspace
`get_board`	Get details of a specific board
`create_board`	Create a new board
`update_board`	Update board name or description
`delete_board`	Delete a board
`list_lists`	List all lists in a board
`create_list`	Create a new list
`update_list`	Update list name or position
`delete_list`	Delete a list
`list_cards`	List cards in a board or list
`get_card`	Get full details of a card
`create_card`	Create a new card
`update_card`	Update card fields
`move_card`	Move a card to a different list
`archive_card`	Archive a card
`delete_card`	Permanently delete a card

Card Fields

When creating or updating a card, the following fields are available:

title, description
type — Task, Bug, Feature, Meeting, Music, Video, Image, Audio, Text
priority, severity, effort
tags, due_date, responsible, job_number
rating (1–5)
links, checklist
prompt, ai_tool — AI-specific fields

Environment Variables

# Required: URL of your Pix3lBoard instance
PIX3LBOARD_URL=http://localhost:3000

# Authentication — choose ONE option:

# Option 1 (preferred): API Key
PIX3LBOARD_API_KEY=pk_live_xxxxxxxxxxxx

# Option 2 (fallback): Email + Password
PIX3LBOARD_EMAIL=admin@example.com
PIX3LBOARD_PASSWORD=YourPassword123

# Workspace filtering (comma-separated workspace IDs, visible in the board URL)
WORKSPACE_IDS=ws_abc123,ws_def456

# Docker variant only
PORT=3010

# Logging
LOG_LEVEL=info

Tip: Your workspace ID appears in the Pix3lBoard URL when you open a board: https://your-board.vercel.app/workspace/ws_abc123/...

Installation

Option 1: Deploy on Vercel (recommended)

Fork or clone the repository and push to GitHub
Import in Vercel, set Framework Preset to Next.js
Add environment variables: PIX3LBOARD_URL, PIX3LBOARD_API_KEY, WORKSPACE_IDS
Deploy
Register in Claude Code:

claude mcp add --transport http --scope user pix3lboard -- https://<your-deployment>.vercel.app/api/mcp

Or add to Claude.ai via Settings → Integrations → Add MCP Server.

Option 2: Docker (with pix3ltools-deploy)

git clone https://github.com/Pix3ltools-lab/pix3ltools-deploy.git
cd pix3ltools-deploy
./setup.sh

After setup, Pix3lMCP runs on http://localhost:3010. Register it in Claude Code:

claude mcp add --transport http --scope user pix3lboard -- http://localhost:3010/mcp

Option 3: Local development (Vercel variant)

git clone https://github.com/Pix3ltools-lab/pix3lmcp.git
cd pix3lmcp
npm install
npm run dev:vercel    # http://localhost:3000/api/mcp

claude mcp add --transport http --scope user pix3lboard -- http://localhost:3000/api/mcp

Option 4: Local development (Docker/Express variant)

npm run dev:docker    # http://localhost:3010/mcp

claude mcp add --transport http --scope user pix3lboard -- http://localhost:3010/mcp

Usage Examples

Once registered, you can ask Claude Code:

“Show me all my boards in workspace ws_abc123”
“Create a new card called ‘Midjourney portrait’ in the ‘In Progress’ list”
“Move the card ‘Logo design’ to the ‘Done’ list”
“Add a prompt to card ‘Hero image’: a cinematic portrait, golden hour, —ar 16:9 —v 6”
“Archive all cards in the ‘Done’ list of board ‘Project X‘“

Architecture

src/
├── auth.ts          # API Key or Email/Password authentication
├── api.ts           # HTTP client for Pix3lBoard REST API v1
└── tools/
    ├── boards.ts    # Board management tools
    ├── lists.ts     # List management tools
    ├── cards.ts     # Card management tools
    └── index.ts     # Tool registration + list_workspaces

app/api/mcp/         # Vercel: Next.js API route (Streamable HTTP)
server.ts            # Docker: Express entry point

npm Scripts

npm run dev:vercel    # Next.js dev server (Vercel variant)
npm run dev:docker    # tsx watch (Docker/Express variant)
npm run build:vercel  # Next.js production build
npm run build:docker  # esbuild → dist/server.js
npm run build         # Build both variants
npm run start:vercel  # Start Next.js production server
npm run start:docker  # Start Express production server
npm run lint          # ESLint