Claude Code Plugin for protoLabs

Setup guide for the Claude Code plugin and MCP server. For commands and examples, see Plugin Commands. For the full MCP tool catalog, see MCP Tools Reference.

Quick Start

Option 1: Install from GitHub (Recommended)

Install directly from the GitHub repository:

bash

# 1. Clone the repo (needed for the MCP server binary)
git clone https://github.com/protoLabsAI/protomaker.git
cd protomaker

# 2. Add the protoLabs plugin marketplace
claude plugin marketplace add https://github.com/protoLabsAI/protomaker/tree/main/packages/mcp-server/plugins

# 3. Install the plugin
claude plugin install protolabs

# 4. Configure the plugin — AUTOMAKER_ROOT is required
#    The plugin needs to know where your local clone lives
PLUGIN_DIR=~/.claude/plugins/protolabs
cp "$PLUGIN_DIR/.env.example" "$PLUGIN_DIR/.env"
# Edit .env and set AUTOMAKER_ROOT to the absolute path of this repo
echo "AUTOMAKER_ROOT=$(pwd)" >> "$PLUGIN_DIR/.env"
echo "AUTOMAKER_API_KEY=your-dev-key-2026" >> "$PLUGIN_DIR/.env"

# 5. Start protoLabs server (in a separate terminal)
npm install
npm run dev:web

# 6. Verify it works
claude
> /board

Option 2: Install from Local Clone (Development)

For developers working on protoLabs:

bash

# 1. Clone and install protoLabs
git clone https://github.com/protoLabsAI/protomaker.git
cd protomaker
npm install

# 2. Build the MCP server
npm run build:packages

# 3. Add the plugin marketplace and install
claude plugin marketplace add $(pwd)/packages/mcp-server/plugins
claude plugin install protolabs

# 4. Configure the plugin — AUTOMAKER_ROOT is required
PLUGIN_DIR=~/.claude/plugins/protolabs
cp "$PLUGIN_DIR/.env.example" "$PLUGIN_DIR/.env"
echo "AUTOMAKER_ROOT=$(pwd)" > "$PLUGIN_DIR/.env"
echo "AUTOMAKER_API_KEY=your-dev-key-2026" >> "$PLUGIN_DIR/.env"

# 5. Start protoLabs server (in a separate terminal)
npm run dev:web

# 6. Verify it works
claude
> /board

That's it! You now have access to ~170 MCP tools and slash commands for managing your Kanban board directly from Claude Code.

What You Can Do

Command	Description
`/welcome`	User onboarding and status refresh
`/board`	View and manage your Kanban board
`/auto-mode`	Start/stop autonomous feature processing
`/orchestrate`	Manage feature dependencies
`/context`	Manage AI agent context files
`/plan-project`	Full project orchestration pipeline
`/ship`	Stage, commit, push, create PR, auto-merge
`/headsdown`	Deep work mode — process features autonomously
`/due-diligence`	Validate approaches with evidence-based research
`/deep-research`	Research codebase before planning
`/sparc-prd`	Create a SPARC-style PRD
`/improve-prompts`	Analyze and improve LLM prompts
`/ava`	Autonomous operator with multi-project delegation
`/setuplab`	7-phase project onboarding pipeline
`/update-plugin`	Guided plugin version upgrade

See Plugin Commands for full command reference and examples.

Overview

The protoLabs Claude Code plugin enables you to:

Manage Features: Create, update, and track features on your Kanban board
Control Agents: Start, stop, and monitor AI agents working on features
Orchestrate Workflows: Set up dependencies and run auto-mode for autonomous processing
Configure Context: Manage context files that guide AI agent behavior

The plugin consists of:

MCP Server (@protolabsai/mcp-server) - Exposes protoLabs's API via Model Context Protocol
Claude Plugin (packages/mcp-server/plugins/automaker) - Provides slash commands and subagents

Installation

Prerequisites

protoLabs server running (npm run dev from the protomaker root directory)
Claude Code CLI installed and authenticated
Node.js 22+

Step 1: Start protoLabs

The server uses protoLabs_studio_key as the default API key. To override, set it in your .env or shell:

bash

AUTOMAKER_API_KEY=your-custom-key npm run dev

Step 2: Build the MCP Server (if not already built)

bash

cd packages/mcp-server
npm run build

Step 2.5: Configure the Plugin Environment

The plugin requires an environment file to locate your local protomaker clone.

bash

# Find where Claude Code installed the plugin
PLUGIN_DIR=~/.claude/plugins/protolabs

# Create the .env from the example template
cp "$PLUGIN_DIR/.env.example" "$PLUGIN_DIR/.env"

Open $PLUGIN_DIR/.env and set AUTOMAKER_ROOT to the absolute path of your protomaker repository root:

AUTOMAKER_ROOT=/absolute/path/to/your/protomaker
AUTOMAKER_API_KEY=your-dev-key-2026

Note: AUTOMAKER_ROOT must be an absolute path. Relative paths and ~ are not expanded by the plugin loader.

Step 3: Install the Plugin

Option A: Add via Marketplace (Recommended)

bash

# Add the local marketplace
claude plugin marketplace add /path/to/protomaker/packages/mcp-server/plugins

# Install the plugin
claude plugin install protolabs

Option B: Symlink Directly

bash

# Create symlink to Claude's plugins directory
ln -s /path/to/protomaker/packages/mcp-server/plugins/automaker ~/.claude/plugins/protolabs

Step 4: Restart Claude Code

After installation, restart Claude Code to load the plugin:

bash

claude

Step 5: Verify Installation

Test that the plugin is working:

/board

You should see your Kanban board or a message to start the protoLabs server.

Configuration

Environment Variables

Configure these in the plugin's .env file (~/.claude/plugins/protolabs/.env). Copy .env.example in that directory to get started.

Variable	Description	Default
`AUTOMAKER_ROOT`	Absolute path to your local protomaker repo clone	(required)
`AUTOMAKER_API_KEY`	API key matching the one the protoLabs server is started with	(required)
`AUTOMAKER_API_URL`	protoLabs API base URL	`http://localhost:3008`
`GH_TOKEN`	GitHub token for PR operations (`gh auth token` to get it)	(optional)
`DISCORD_BOT_TOKEN`	Discord bot token for Discord MCP tools	(optional)
`CONTEXT7_API_KEY`	Context7 API key for documentation lookup	(optional)
`ENABLE_TOOL_SEARCH`	Tool search mode (`auto:N` to limit active tools)	`auto:10`

AUTOMAKER_ROOT is the most common cause of a new install failing silently. Without it, the plugin cannot locate the MCP server executable and no tools will be available.

Plugin Configuration

The plugin configuration is in packages/mcp-server/plugins/automaker/.claude-plugin/plugin.json:

json

{
  "name": "automaker",
  "description": "protoLabs Studio — AI Development Studio. Manage Kanban boards, AI agents, and feature orchestration.",
  "version": "1.1.1",
  "mcpServers": {
    "automaker": {
      "command": "bash",
      "args": ["${AUTOMAKER_ROOT}/packages/mcp-server/plugins/automaker/hooks/start-mcp.sh"],
      "env": {
        "AUTOMAKER_API_URL": "http://localhost:3008",
        "AUTOMAKER_API_KEY": "${AUTOMAKER_API_KEY}",
        "GH_TOKEN": "${GH_TOKEN}",
        "ENABLE_TOOL_SEARCH": "auto:10"
      }
    }
  }
}

The MCP server is launched via start-mcp.sh which handles path resolution and env loading automatically.

Docker Deployment

For production deployments using Docker, follow these steps to configure the MCP plugin to communicate with a containerized protoLabs server.

Prerequisites

Docker and Docker Compose installed
protoLabs repository cloned
Claude Code CLI installed

Step 1: Configure docker-compose.override.yml

Create docker-compose.override.yml in the protomaker root directory:

yaml

# protoLabs Production Override
# Mounts host project directories for development work

services:
  server:
    environment:
      - ALLOWED_ROOT_DIRECTORY=/home/youruser/dev
      - AUTOMAKER_API_KEY=your-secure-api-key

    volumes:
      # CRITICAL: Container path must match host path exactly
      - /home/youruser/dev:/home/youruser/dev:rw
      - automaker-data:/data
      - automaker-claude-config:/home/automaker/.claude

    deploy:
      resources:
        limits:
          memory: 8G
        reservations:
          memory: 4G

Path Mapping Note: The MCP plugin runs on the host and passes paths like /home/youruser/dev/myproject to the containerized server. The container must have the exact same path available.

Step 2: Configure .env File

Create or update .env in the protomaker root directory:

bash

AUTOMAKER_API_KEY=your-secure-api-key
UID=1000
GID=1000
HOST=0.0.0.0
PORT=3008
CORS_ORIGIN=http://localhost:3007
ALLOWED_ROOT_DIRECTORY=/home/youruser/dev
VITE_HOSTNAME=localhost

Step 3: Update Plugin Environment

Set AUTOMAKER_ROOT in the plugin .env to point at your protomaker clone. The start-mcp.sh launcher handles path resolution from there:

bash

PLUGIN_DIR=~/.claude/plugins/protolabs
echo "AUTOMAKER_ROOT=/absolute/path/to/protomaker" > "$PLUGIN_DIR/.env"
echo "AUTOMAKER_API_KEY=your-secure-api-key" >> "$PLUGIN_DIR/.env"

Step 4: Build and Start

bash

npm run build:packages
docker compose up -d
curl http://localhost:3008/api/health

Step 5: Install/Reinstall Plugin

bash

claude plugin marketplace add /path/to/protomaker/packages/mcp-server/plugins
claude plugin install protolabs

Memory Allocation Guidelines

Use Case	Memory Limit	Reservation
Light (1-2 concurrent agents)	4G	2G
Medium (3-5 concurrent agents)	8G	4G
Heavy (6+ concurrent agents)	16G	8G

Working with Multiple Projects

With the path mapping above, you can use protoLabs with any project under your ALLOWED_ROOT_DIRECTORY:

bash

# Project A
cd /home/youruser/dev/project-a
claude
> /board

# Project B (different terminal)
cd /home/youruser/dev/project-b
claude
> /board

Each project maintains its own .automaker/ directory with independent features, settings, and context files.

Troubleshooting

AUTOMAKER_ROOT Not Set

Symptom: No MCP tools appear in Claude Code. Session start prints AUTOMAKER_ROOT is not set.

Fix:

bash

PLUGIN_DIR=~/.claude/plugins/protolabs
cp "$PLUGIN_DIR/.env.example" "$PLUGIN_DIR/.env"
# Open .env and set AUTOMAKER_ROOT to the absolute path of your protomaker clone
echo "AUTOMAKER_ROOT=/absolute/path/to/protomaker" >> "$PLUGIN_DIR/.env"
echo "AUTOMAKER_API_KEY=your-dev-key-2026" >> "$PLUGIN_DIR/.env"

Restart Claude Code after editing .env.

MCP Server Binary Not Found

Symptom: AUTOMAKER_ROOT is set but session start prints MCP server binary not found.

Fix: The MCP server has not been built yet.

bash

cd "$AUTOMAKER_ROOT"
npm run build:packages

Then restart Claude Code.

Plugin Not Loading (General)

If neither of the above applies:

Verify the plugin is installed: ls ~/.claude/plugins/
Check ~/.claude/plugins/protolabs/.env exists and contains AUTOMAKER_ROOT
Confirm AUTOMAKER_ROOT points to the correct repo: ls "$AUTOMAKER_ROOT/packages/mcp-server/dist/index.js"
Restart Claude Code

Connection Errors

Ensure protoLabs server is running: npm run dev
Check the API URL matches: curl http://localhost:3008/api/health
Verify the API key: echo $AUTOMAKER_API_KEY

Authentication Errors

Ensure AUTOMAKER_API_KEY is set in both protoLabs server and plugin configuration
The keys must match exactly

Tools Not Available

Check the health endpoint: /board — if it fails, start protoLabs
Verify MCP tools are loaded: mcp__protolabs__health_check()

Feature Dependencies Not Working

Ensure features exist before setting dependencies: /board
Check for circular dependencies: /orchestrate

GitHub Operations Fail

PR-related tools require GH_TOKEN:

Get your token: gh auth token
Add to .env: GH_TOKEN=gho_xxxxx
Add to plugin.json env: "GH_TOKEN": "${GH_TOKEN}"
Restart the server and reload the plugin

Docker-Specific Issues

"Path not found" or "Permission denied":

Verify path mapping in docker-compose.override.yml — host and container paths must match
Check ALLOWED_ROOT_DIRECTORY includes your project

"Unauthorized" or API key errors:

Ensure API key matches in both docker-compose.override.yml and plugin.json
Restart containers after changing keys: docker compose down && docker compose up -d

"Cannot find module" in MCP server:

Ensure plugin.json uses absolute path in args
Rebuild: npm run build:packages
Reinstall plugin: claude plugin install protolabs

Container memory issues:

Check usage: docker stats automaker-server-1
Increase limits in docker-compose.override.yml

Third-party MCP Docker images (ARM64 vs AMD64):

Some MCP Docker images are built only for ARM64. On Linux/AMD64, build locally:

bash

git clone https://github.com/SaseQ/discord-mcp /tmp/discord-mcp
cd /tmp/discord-mcp
docker build --platform linux/amd64 -t discord-mcp:amd64 .

Plugin Quickstart — 5-minute setup guide
Plugin Deep Dive — Architecture, hooks, tools, and extension points
Plugin Commands — Commands reference, subagents, examples
MCP Tools Reference — Full MCP tool catalog
Context System — Best practices for context files
Getting Started — Main project documentation

Claude Code Plugin for protoLabs ​

Quick Start ​

Option 1: Install from GitHub (Recommended) ​

Option 2: Install from Local Clone (Development) ​

What You Can Do ​

Overview ​

Installation ​

Prerequisites ​

Step 1: Start protoLabs ​

Step 2: Build the MCP Server (if not already built) ​

Step 2.5: Configure the Plugin Environment ​

Step 3: Install the Plugin ​

Step 4: Restart Claude Code ​

Step 5: Verify Installation ​

Configuration ​

Environment Variables ​

Plugin Configuration ​

Docker Deployment ​

Prerequisites ​

Step 1: Configure docker-compose.override.yml ​

Step 2: Configure .env File ​

Step 3: Update Plugin Environment ​

Step 4: Build and Start ​

Step 5: Install/Reinstall Plugin ​

Memory Allocation Guidelines ​

Working with Multiple Projects ​

Troubleshooting ​

AUTOMAKER_ROOT Not Set ​

MCP Server Binary Not Found ​

Plugin Not Loading (General) ​

Connection Errors ​

Authentication Errors ​

Tools Not Available ​

Feature Dependencies Not Working ​

GitHub Operations Fail ​

Docker-Specific Issues ​

Related Documentation ​

Claude Code Plugin for protoLabs

Quick Start

Option 1: Install from GitHub (Recommended)

Option 2: Install from Local Clone (Development)

What You Can Do

Overview

Installation

Prerequisites

Step 1: Start protoLabs

Step 2: Build the MCP Server (if not already built)

Step 2.5: Configure the Plugin Environment

Step 3: Install the Plugin

Step 4: Restart Claude Code

Step 5: Verify Installation

Configuration

Environment Variables

Plugin Configuration

Docker Deployment

Prerequisites

Step 1: Configure docker-compose.override.yml

Step 2: Configure .env File

Step 3: Update Plugin Environment

Step 4: Build and Start

Step 5: Install/Reinstall Plugin

Memory Allocation Guidelines

Working with Multiple Projects

Troubleshooting

AUTOMAKER_ROOT Not Set

MCP Server Binary Not Found

Plugin Not Loading (General)

Connection Errors

Authentication Errors

Tools Not Available

Feature Dependencies Not Working

GitHub Operations Fail

Docker-Specific Issues

Related Documentation