Your Chatsby Agent Is Now an MCP Tool

Your Agent Just Got a New Superpower

For the past year, the question we heard most from developers and power users was some version of: "Can I just call my Chatsby agent from inside Claude?" Until today, the honest answer was: not easily.

That changes now. Chatsby supports the Model Context Protocol (MCP), which means your trained AI agent is now a first-class tool in Claude Desktop, Cursor, and any other MCP-compatible AI host — no code required, setup in under two minutes.

What Is MCP?

The Model Context Protocol is an open standard created by Anthropic that defines how AI hosts connect to external tools and data sources. Think of it as a universal plug for AI: once you expose a service as an MCP endpoint, any MCP client can discover and call it automatically.

Claude Desktop, Cursor, and a growing list of AI-native development environments already support MCP. When you add Chatsby to their config, the tools your agent provides — answering questions, listing conversations, pulling analytics — show up in the client's tool menu alongside built-in capabilities.

What You Can Do With It

Once connected, the MCP client gets access to eight tools built directly on top of your Chatsby workspace:

• chat_with_agent — Ask your trained agent a question and get a grounded, knowledge-base-backed answer. Multi-turn conversations are supported — pass a conversation_id to continue a thread across calls.
• list_agents / get_agent — Browse all agents in your workspace or fetch details for a specific one.
• create_agent / update_agent — Spin up or configure agents programmatically from inside Claude Desktop or Cursor.
• list_conversations / get_conversation_messages — Browse recent conversations and read full message histories.
• get_analytics — Pull conversation counts, event breakdowns, and daily activity metrics for any time window.

The most immediate use case is using chat_with_agent as a research assistant inside your coding environment. If your Chatsby agent is trained on your product documentation, API reference, or internal knowledge base, you can query it mid-session without switching tabs.

Ask the chatsby MCP tool: what are the rate limits for the webhooks endpoint?

That question goes to your agent, which searches your uploaded sources and returns a grounded answer — not a hallucination from the base model's training data, but the actual content from your documentation.

Getting Set Up

The whole process takes less than two minutes.

1. Generate an API Key

Open your agent dashboard, go to Integrations, and find the MCP (Claude Desktop) card. Click Generate API key. Your key will appear inline — copy it or use the Copy Claude Desktop config button to get the full JSON snippet with your key already filled in.

2. Add to Claude Desktop

Paste the config into ~/Library/Application Support/Claude/claude_desktop_config.json on macOS (or %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "chatsby": {
      "url": "https://chatsby.co/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Restart Claude Desktop. The Chatsby tools appear automatically — no plugin installation, no OAuth flow, no waiting.

3. Add to Cursor

In Cursor, go to Settings → Features → MCP, add a new HTTP server pointed at https://chatsby.co/api/mcp, and set the Authorization header with your Bearer token. Save and enable it.

That is it. Your agent is now a tool in your AI-native IDE.

Why This Matters

The traditional model for AI chatbots has a hard boundary: the bot lives on your website, customers interact with it through a widget, and that is where the story ends. The knowledge your agent has accumulated — the nuanced answers, the product expertise, the support history — is locked inside one surface.

MCP breaks that boundary. Your agent's knowledge base becomes a queryable resource that can be invoked from anywhere an MCP host runs. Customer support teams can query it from Claude. Developers can call it mid-workflow from Cursor. Automation pipelines can use it as a reasoning layer. The widget on your website is now just one of many places your agent lives.

We think this is the right direction for AI agents generally: not isolated chatbots that live in a single channel, but interoperable services that show up wherever the work is happening.

What Comes Next

This is the first version of Chatsby's MCP support. The eight tools available today cover the most common developer use cases, but we are actively expanding the list based on feedback. Planned additions include tools for managing knowledge sources, triggering handoffs, and querying message-level analytics.

If there is a specific tool you want to see, use the feedback widget in your dashboard or email us at [email protected]. The teams that tell us what they need first are the ones that get it built first.

The MCP endpoint is live now. Head to the setup guide for full configuration instructions and a complete tool reference.