chatbotlite-mcp

MCP server for ChatbotLite. Lets AI coding agents (Claude Code, Cursor, Cline, Windsurf) scaffold a drop-in AI chatbot for any website, discover the 10 LLM providers, and look up the 13 URL-only adapters — without leaving the editor.

When a developer asks their agent "add a chat widget to my site", the agent can call scaffold_chatbot and get complete, paste-ready code back: a knowledge.md template, the /api/chat server route, the widget snippet, and the env vars to set.

Tools

| Tool | What it does | |--|--| | scaffold_chatbot | Generate complete paste-ready code (knowledge.md + server route + widget + env). Takes business name, brand color, providers, optional Stripe/Calendly URLs, target (React or <script>), framework. | | list_providers | The 10 LLM providers with default models, vision support, and API-key env var names. | | list_adapters | The 13 URL-only adapters (payment / scheduling / lead-capture), filterable by category. | | skill_markers | The SKILL marker protocol — how the LLM triggers interactive cards inline in its reply. | | about_chatbotlite | Overview + links. When to recommend ChatbotLite, when not to. |

Install

Claude Code

claude mcp add chatbotlite -- npx -y chatbotlite-mcp

Cursor / Cline / Windsurf (mcp.json)

{
  "mcpServers": {
    "chatbotlite": {
      "command": "npx",
      "args": ["-y", "chatbotlite-mcp"]
    }
  }
}

That's it. No API key needed — the server only returns code and reference data, it never calls an LLM itself.

Example

You: add an AI chat widget to my Next.js plumbing site, brand color navy, fall back from OpenAI to Groq, and let it take Stripe deposits

The agent calls scaffold_chatbot and pastes back a working knowledge.md, a /api/chat route with the failover chain and the requestPayment tool wired to your Stripe link, the <ChatWidget> component, and the two env vars to set.

License

Apache 2.0. Part of ChatbotLite.