MCP Server Integration

ChatReact implements the Model Context Protocol (MCP), allowing you to connect your ChatReact account to AI assistants like Claude Desktop, Cursor, and other MCP-compatible tools.

What is MCP?

The Model Context Protocol is an open standard that enables AI assistants to interact with external services and data sources. With ChatReact's MCP integration, you can:

• Manage chatbots — create, configure (persona, languages, widget, behavior, restrictions), embed and delete chatbots
• Build the knowledge base — add documents, crawl websites, manage crawled pages and core instructions
• Manage FAQs — FAQs, categories, suggestions, translations and the FAQ widget
• Review chats — rate conversations, assign confidence categories, run the Improvement Agent
• Work with leads and forms — leads, lead categories, capture settings, contact forms and submissions
• Act as a live chat agent — set your availability, take over waiting chats, reply and close sessions
• Administer your company — profile, brand colors, team members and invitations
• Use the Email Assistant — mailbox connections, categories, contacts and activity
• Read analytics and notifications

Everything you can do in the dashboard is available through MCP (and the REST API); only browser-bound actions such as payments, mailbox OAuth, passkeys, account deletion and API key management stay in the dashboard.

Getting Started

1. Generate an API Key

1. Navigate to your Dashboard → Company → API tab
2. Click Create API Key
3. Give your key a descriptive name (e.g., "Claude Desktop") and select at least one permission
4. Copy the generated key (starts with cr_)

2. Configure Your MCP Client

The ChatReact MCP server is hosted at https://www.chatreact.ai/api/mcp and communicates over HTTP (JSON-RPC). Authenticate with an Authorization: Bearer header — an X-API-Key header with the same key is also accepted.

Cursor IDE

Cursor supports remote MCP servers directly. Add to your Cursor MCP settings:

{
  "mcpServers": {
    "chatreact": {
      "url": "https://www.chatreact.ai/api/mcp",
      "headers": {
        "Authorization": "Bearer cr_your_api_key_here"
      }
    }
  }
}

Claude Desktop

Claude Desktop's configuration file supports command-based servers, so use the standard mcp-remote bridge to connect to the hosted endpoint:

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

{
  "mcpServers": {
    "chatreact": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://www.chatreact.ai/api/mcp",
        "--header",
        "X-API-Key:cr_your_api_key_here"
      ]
    }
  }
}

3. Restart Your AI Assistant

After saving the configuration, restart Claude Desktop or Cursor to establish the connection.

Available Tools

ChatReact MCP exposes the tools below, grouped by area. The Scope column shows the API key permission a tool requires (any = every valid key). Everything the dashboard can do is available here as well; tools run with the same permissions, plan limits and side effects as the dashboard.

Account

Company & Team

Chatbot Management

Knowledge Base

FAQ Management

Chat History & Feedback

Improvement Agent

Live Chat

Leads

Forms

Email Assistant

Analytics

Example Workflows

Adding a Core Instruction

Ask your AI assistant:

"Add a core instruction to my 'Support Bot' chatbot: we offer free shipping on orders over €50"

The assistant will use chatbot_add_instruction with appropriate parameters.

Reviewing Improvement Suggestions

"Show me all pending improvement proposals for my chatbot and explain what changes they suggest"

The assistant will:

1. Use improvement_proposals_pending to fetch proposals
2. Analyze and summarize each suggestion
3. Approve or reject them on your behalf with improvement_proposal_approve or improvement_proposal_reject

Analyzing Chat Performance

"Give me a weekly report of chat volume and customer satisfaction for my chatbot"

The assistant will use analytics_overview and analytics_chat_volume to compile a comprehensive report.

Importing a Website

"Scan https://example.com, import the product and pricing pages into my 'Support Bot' and tell me when the crawl is done"

The assistant will call crawler_preview_scan, pick the relevant pages, start the import with crawler_start and poll crawler_job_get until the job has finished.

Reviewing Leads and Form Submissions

"Show me the new leads from this week, assign the qualified ones to Anna and approve the pending contact form submissions"

The assistant combines lead_list, lead_assign, form_submissions_list and form_submission_update_status.

Taking Over a Live Chat

"Set me available for live chat and take the oldest waiting conversation"

The assistant uses live_chat_set_availability, live_chat_queue and live_chat_session_assign; replies go through live_chat_message_send.

Security

• API keys are scoped to your company and to the permission scopes you select when creating them (see the Scope column above)
• Keys execute with the permissions of the company owner; plan limits and feature gates apply exactly as in the dashboard
• Keys can be revoked at any time from the dashboard
• API access is available on paid plans — see our Pricing page

Troubleshooting

Connection Failed

1. Verify your API key is correct
2. Check that the key hasn't been revoked
3. Make sure your plan includes API access
4. Restart your MCP client

Tool Not Found

Restart your MCP client after changing the configuration so it reloads the available tools. Tool names are lowercase with underscores (e.g., faq_list).

Need help? Contact our support team →