AI智能体管理MCP服务器

AnroAgents MCP Server

MCP (Model Context Protocol) server for managing AI agents on the AnroAgents platform. Control your agents directly from Claude, Cursor, Windsurf, Gemini CLI and other MCP clients — and from ChatGPT through the same API.

The server runs on your own computer: your AI client starts it with npx, so it needs Node.js 18 or newer. It works in desktop apps, CLIs and code editors, not in web chats such as claude.ai in the browser.

Features

Quick Start

1. Get a Personal Access Token

1. Log in to app.anroagents.com
2. Go to Settings > API Tokens
3. Click Create Token, give it a name (e.g., "Claude Desktop")
4. Copy the token (it starts with pat_ and is shown only once)

2. Configure your AI client

Choose your client below and follow the instructions.

Setup by Client

Claude Desktop

Open Settings > Developer > Edit Config, or edit the file directly:

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

{
  "mcpServers": {
    "anroagents": {
      "command": "npx",
      "args": ["-y", "@anroagents/mcp-server"],
      "env": {
        "ANROAGENTS_TOKEN": "pat_your_token_here"
      }
    }
  }
}

Quit Claude Desktop completely (Cmd+Q on macOS) and open it again — closing the window is not enough.

Claude Code (CLI)

Claude Code registers MCP servers with its own command — it does not read mcpServers from settings.json:

claude mcp add anroagents --scope user -e ANROAGENTS_TOKEN=pat_your_token_here -- npx -y @anroagents/mcp-server

--scope user makes it available in every project. Check with claude mcp list. To share it with a team through the repository instead, put the same mcpServers block as for Claude Desktop into .mcp.json at the project root (and keep the token out of git).

Cursor

Open Settings > MCP > Add new MCP server and configure:

• Name: anroagents
• Command: npx -y @anroagents/mcp-server
• Environment: ANROAGENTS_TOKEN=pat_your_token_here

Windsurf

Open Settings > MCP > Add Server:

{
  "anroagents": {
    "command": "npx",
    "args": ["-y", "@anroagents/mcp-server"],
    "env": {
      "ANROAGENTS_TOKEN": "pat_your_token_here"
    }
  }
}

VS Code (with Continue or Copilot Chat MCP)

Add to your MCP configuration:

{
  "servers": {
    "anroagents": {
      "command": "npx",
      "args": ["-y", "@anroagents/mcp-server"],
      "env": {
        "ANROAGENTS_TOKEN": "pat_your_token_here"
      }
    }
  }
}

Gemini CLI

Add to ~/.gemini/settings.json:

{
  "mcpServers": {
    "anroagents": {
      "command": "npx",
      "args": ["-y", "@anroagents/mcp-server"],
      "env": {
        "ANROAGENTS_TOKEN": "pat_your_token_here"
      }
    }
  }
}

ChatGPT (Custom GPT Actions)

ChatGPT cannot start a local MCP server like this one. Instead, point a Custom GPT at the same API through its OpenAPI spec:

1. Create a Custom GPT at chatgpt.com
2. Go to Configure > Actions > Import from URL
3. Enter: https://api.anroagents.com/mcp/openapi.json
4. Set Authentication: API Key, auth type Bearer, value pat_your_token_here
5. Save and test

Environment Variables

Troubleshooting

If your client shows the server as failed or disconnected, the cause is almost always on your computer, not in your AnroAgents account:

• npx: command not found / spawn npx ENOENT — Node.js is missing or the app can't see it. Install Node.js 18+ from nodejs.org. If you installed it with nvm or Homebrew, GUI apps may not see it: set "command" to the full path from which npx.
• ANROAGENTS_TOKEN environment variable is required — the env block is missing or misspelled in the config.
• HTTP 401 or HTTP 403 on every tool call — the token was deleted or mistyped. A token is shown only once; create a new one in Settings > API Tokens and paste it into the config.
• Logs — Claude Desktop writes the reason to ~/Library/Logs/Claude/mcp-server-anroagents.log (macOS) or %APPDATA%\Claude\logs\mcp-server-anroagents.log (Windows).

Connection instructions are also always available in the dashboard under Settings > API Tokens > How to connect an AI assistant.

Tool Reference

list-agents

List all agents owned by the authenticated user.

Parameters:

• status (optional): Filter by status — draft, review, active, or disabled

get-agent

Get full details of a specific agent.

Parameters:

• agentId (required): The agent's unique ID

create-agent

Create a new AI agent. The agent starts in draft status.

Parameters:

• businessName (required): Name of the business
• businessDescription (optional): Description of what the business does
• defaultGreeting (optional): Greeting message shown to visitors
• defaultLanguage (optional): Language code (default: en)
• communicationTone (optional): formal, friendly, or neutral
• businessHours (optional): e.g., "Mon-Fri 9am-5pm"
• widgetTitle (optional): Title shown in the chat widget
• escalationRules (optional): When to escalate to a human
• restrictions (optional): Topics the agent should avoid
• bookingUrl (optional): Online booking URL
• attachmentsEnabled (optional): Let visitors send photos (the agent sees them) and files in chat — Starter plan or higher
• chatModel (optional): standard (default) or advanced — the advanced model writes better sales conversations; each advanced reply counts as 15 messages against the monthly plan limit. Starter plan or higher
• services (optional): Array of { name, price, description }
• contacts (optional): { website, email, phone, address }
• faq (optional): Array of { question, answer }

update-agent

Update an existing agent. Only provide fields you want to change.

Parameters: Same as create-agent, plus agentId (required).

delete-agent

Delete an agent. Only agents in draft status can be deleted.

Parameters:

• agentId (required): The agent's unique ID

submit-agent-for-review

Submit an agent for review and catalog listing. Requires a paid plan and active subscription.

Parameters:

• agentId (required): The agent's unique ID

toggle-agent

Enable or disable an active agent.

Parameters:

• agentId (required): The agent's unique ID

set-agent-logo

Upload a logo image and attach it to an agent. Provide either a local file or a public image URL. Accepts PNG, JPEG, WebP, or SVG up to 10 MB. The image is uploaded to secure storage and the agent is updated in one step.

Parameters:

• agentId (required): The agent's unique ID
• filePath (optional): Absolute path to a local image file. Use this or imageUrl.
• imageUrl (optional): Public URL of an image to fetch. Use this or filePath.

list-knowledge

List an agent's knowledge base sources — uploaded files, pages and whole websites — with their status (pending, processing, ready, error), page counts for websites, and the plan limits.

Parameters:

• agentId (required): The agent's unique ID

add-website-knowledge

Add a whole website to the agent's knowledge base. The pages are read in the background within the plan's page and character budgets, and the site becomes one knowledge source. It starts as pending; use list-knowledge to see when it is ready. Addresses on internal networks are refused, and each site can be added once — use resync-knowledge to read it again.

Parameters:

• agentId (required): The agent's unique ID
• url (required): Website address, e.g. https://example.com or example.com

resync-knowledge

Read a knowledge source again: a website is crawled again from the web, a file is re-indexed.

Parameters:

• agentId (required): The agent's unique ID
• fileId (required): The source ID from list-knowledge

get-embed-code

Get the HTML snippet to embed the chat widget on any website.

Parameters:

• agentId (required): The agent's unique ID

regenerate-agent-key

Regenerate the agent's API key. The old key stops working immediately.

Parameters:

• agentId (required): The agent's unique ID

Usage Examples

Once configured, you can ask your AI assistant things like:

• "List my AnroAgents agents"
• "Create a new agent for a pizza restaurant called Mario's Pizza"
• "Update my agent's business hours to Mon-Sat 11am-10pm"
• "Add a FAQ to my agent: Q: Do you deliver? A: Yes, within 5 miles"
• "What's the embed code for my Mario's Pizza agent?"
• "Set the logo for my Mario's Pizza agent from ~/Downloads/logo.png"
• "Teach my Mario's Pizza agent everything on mariospizza.com"
• "Let visitors send photos to my carpentry agent"
• "Disable my test agent"
• "Delete my draft agent"

Development

# Clone the repository
git clone https://github.com/Anrotech/anroagents-mcp.git
cd anroagents-mcp

# Install dependencies
npm install

# Run locally (requires token)
ANROAGENTS_TOKEN=pat_... npm run dev

# Build for production
npm run build

# Test with MCP Inspector
ANROAGENTS_TOKEN=pat_... npx @modelcontextprotocol/inspector npx tsx src/bin/anroagents-mcp.ts

License

MIT - see LICENSE for details.