Skip to main content

MCP server that connects to Max AI agents

Project description

Max MCP Server

A Model Context Protocol (MCP) server that connects Claude and other AI assistants to AnswerRocket's Max AI platform, enabling access to Max copilots and their skills directly from your AI conversations.

Quick Start for Claude Desktop Users (.dxt install)

Prerequisites

macOS:

  • Install uv via Homebrew: brew install uv
  • If you already have uv installed via another method, uninstall and reinstall with Homebrew:
    # Remove existing uv data and binaries
    uv cache clean
    rm -r "$(uv python dir)"
    rm -r "$(uv tool dir)"
    rm ~/.local/bin/uv ~/.local/bin/uvx
    
    # Install via Homebrew
    brew install uv
    

Windows/Linux:

Installation

  1. Download the .dxt file: mcp-server.dxt

  2. Install Claude Desktop if you haven't already

  3. Double-click the .dxt file to open it with Claude Desktop

  4. Click "Install" when prompted

  5. Configure your Max instance:

    • Max URL: Enter your Max instance URL (e.g., http://localhost:8080 or https://maxai.dev.answerrocket.com)
      • Include the http:// or https:// prefix
      • Do not include any path components
    • SDK Key: Get this from your Max instance frontend
    • Agent ID: This is the copilot ID from Skill Studio
      • Navigate to Skill Studio: https://your-max-instance/apps/system/skill-studio/
      • The URL pattern is: /skill-studio/{COPILOT_ID}/skills/{SKILL_ID}
      • Copy the first UUID (COPILOT_ID)
  6. Click "Save" (keep the setup window open)

  7. Enable the MCP server using the toggle in the upper left

Verification

Navigate back to Claude Desktop's home screen and click the tools icon. You should see "Max MCP" with a list of tools identical to those available in your Max Agent/Copilot.

Note: This feature requires Max instances on version 25.09 or above.

Web-based Usage (Coming Soon)

Support for web-based AI assistants like claude.ai and chatgpt.com is in the works.

Development Setup

Prerequisites

  • Python 3.10+
  • Git
  • Access to a Max instance

Installation

  1. Clone the repository:

    git clone https://github.com/answerrocket/mcp-server.git
    cd mcp-server
    
  2. Install dependencies:

    uv sync
    

Environment Variables

The server supports two modes: local and remote.

Local Mode (Direct API Access)

For development and testing with direct API access to Max:

export MCP_MODE=local
export AR_URL=http://superstoredev.local.answerrocket.com:1234 # or where your instance is located. It does not have to be a local Max instance.
export AR_TOKEN=arc-your-token-here
export COPILOT_ID=your-copilot-uuid
export MCP_TRANSPORT=stdio

Remote Mode (OAuth)

For production deployments with OAuth authentication:

export MCP_MODE=remote
export MCP_TRANSPORT=streamable-http
export MCP_HOST=localhost
export MCP_PORT=9090

Note: In remote mode, the AR_URL is automatically derived from incoming requests, enabling true multi-tenancy.

URL Patterns

When running in remote mode, the server accepts requests at:

http://your-server:port/mcp/agent/{COPILOT_ID}

For example:

  • http://superstoredev.local.answerrocket.com:1234/mcp/agent/0d91262d-e039-43c3-8022-0d285af703d4

This URL pattern is typically routed through your AnswerRocket instance's nginx configuration.

Testing with MCP Inspector

We highly recommend using the MCP Inspector for development and testing:

npx @modelcontextprotocol/inspector

Version 0.16.1 is buggy and does not play well. If that is the case, you should use version 0.16.0

npx @modelcontextprotocol/inspector@0.16.0

Local Development with Claude Desktop

You can test your local development server directly with Claude Desktop by configuring your claude_desktop_config.json:

Configuration File Location

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json

Example Configuration

{
  "mcpServers": {
    "max-mcp-dev": {
      "command": "/path/to/mcp-server/.venv/bin/python",
      "args": [
        "/path/to/mcp-server/mcp_server/__main__.py"
      ],
      "env": {
        "AR_URL": "http://your-max-instance:port",
        "AR_TOKEN": "your-sdk-token",
        "COPILOT_ID": "your-copilot-uuid"
      }
    }
  }
}

Important: Replace the example values with your actual:

  • Max instance URL
  • SDK token
  • Copilot ID

After saving the configuration, restart Claude Desktop to apply the changes.

Remote Deployment with OAuth

For production deployments with OAuth authentication:

Max Instance Configuration

Your Max instance must have OAuth server enabled:

export ENABLE_OAUTH_SERVER=true
./overmind.sh --overlay=superstoredev,ricedemo

Multi-Tenant Testing

To test multiple tenants, configure your Max instance with multiple tenant overlays and ensure your OAuth server supports the tenant domains you want to test.

Server Deployment

  1. Set environment variables:

    export MCP_MODE=remote
    export MCP_TRANSPORT=streamable-http
    export MCP_HOST=0.0.0.0  # or your server IP
    export MCP_PORT=9090
    
  2. Run the server:

    maxai-mcp
    

    or

    uv run python mcp_server/__main__.py
    

OAuth Flow

In remote mode, the server:

  1. Validates OAuth tokens against the requesting domain's introspection endpoint
  2. Dynamically determines the Max instance URL from the request context
  3. Registers tools/skills specific to the requested copilot ID
  4. Supports multiple Max instances simultaneously without configuration changes

API Reference

Supported Transports

  • stdio: For direct integration with Desktop apps
  • streamable-http: && sse For web-based integrations and production deployments

Authentication

  • Local mode: Uses SDK tokens for direct API access
  • Remote mode: Uses OAuth 2.0 with Bearer tokens

Dynamic Tool Registration

The server dynamically loads and registers tools based on:

  • The copilot ID in the request URL
  • Available skills in the specified Max copilot
  • User permissions and authentication context

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

maxai_mcp-0.2.2.tar.gz (15.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

maxai_mcp-0.2.2-py3-none-any.whl (21.9 kB view details)

Uploaded Python 3

File details

Details for the file maxai_mcp-0.2.2.tar.gz.

File metadata

  • Download URL: maxai_mcp-0.2.2.tar.gz
  • Upload date:
  • Size: 15.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.12.9

File hashes

Hashes for maxai_mcp-0.2.2.tar.gz
Algorithm Hash digest
SHA256 7bb844501986e9f575c16bb8cb9b492d4f772a04b05e34e4011c4014fbe3e491
MD5 0525b1a5db3b6600ede491123f56bb07
BLAKE2b-256 2e36183b13ca9765acb86f03b33c7424393c33697b12777ae51c0cc93b56efd1

See more details on using hashes here.

Provenance

The following attestation bundles were made for maxai_mcp-0.2.2.tar.gz:

Publisher: build-and-publish.yml on answerrocket/mcp-server

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file maxai_mcp-0.2.2-py3-none-any.whl.

File metadata

  • Download URL: maxai_mcp-0.2.2-py3-none-any.whl
  • Upload date:
  • Size: 21.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.12.9

File hashes

Hashes for maxai_mcp-0.2.2-py3-none-any.whl
Algorithm Hash digest
SHA256 305e5c53e90e438beb1cac3096a763bbbb7a4afa332f26978b2606e7f1f6906b
MD5 280d63457e47b2dcecfdd770e17fb2e5
BLAKE2b-256 ed973780cb44438851271b32fe52ba7676e6df18093149490205191a6850fe2d

See more details on using hashes here.

Provenance

The following attestation bundles were made for maxai_mcp-0.2.2-py3-none-any.whl:

Publisher: build-and-publish.yml on answerrocket/mcp-server

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page