MCP Server
The MCP Server integration lets you turn any ChatBotKit skillset into a set of tools that AI applications can discover and use through the Model Context Protocol (MCP). Instead of building custom APIs or plugins for every tool, you publish your skillset once and any MCP-compatible client - Claude Desktop, VSCode, GitHub Copilot, ChatGPT, and others - can pick it up automatically.
What You Can Do
With the MCP Server integration, you can:
- Publish Skillsets as Tools: Take any skillset you have built in ChatBotKit and instantly make its abilities available as MCP tools
- Connect to Multiple Clients: Use the same integration with Claude Desktop, VSCode, GitHub Copilot, ChatGPT, and any other MCP-compatible application
- Zero Code Required: No custom server code or infrastructure needed - ChatBotKit hosts the MCP endpoint for you
- Secure Access: Each integration gets its own access token so you control exactly who can call your tools
How It Works
When you create an MCP Server integration and attach a skillset, ChatBotKit generates a hosted MCP endpoint. AI applications connect to this endpoint using the configuration you provide (a URL and an access token). The client discovers all available tools from your skillset and can call them during conversations. ChatBotKit handles the tool execution and returns the results back to the client.
The whole process is transparent to the end user. They interact with their AI application as usual, and the application calls your tools whenever they are relevant to the conversation.
Getting Started
- Create the Integration: Give it a name and click Create
- Select a Skillset: Choose an existing skillset that contains the abilities you want to expose as MCP tools
- Save: Click Save to generate your MCP endpoint URL and access token
- Copy the Configuration: Expand the Installation Instructions section and pick the tab for your client (General, VSCode, ChatGPT, or GitHub Copilot)
- Configure Your Client: Paste the configuration into your client's MCP settings file and restart the application
- Start Using Tools: Your AI application will now discover and use the tools from your skillset
Client Setup
Each MCP client has a slightly different configuration format, but the core information is the same: an endpoint URL and your access token.
Claude Desktop and general clients use a JSON configuration that runs mcp-remote via npx. Paste the provided JSON into your client's MCP settings file and make sure npx is available on your system.
VSCode supports MCP servers natively with an HTTP transport. Add the provided configuration to your VSCode MCP settings and restart the editor. Your skillset tools will appear alongside other MCP tools.
GitHub Copilot also uses an HTTP configuration. Add the JSON to your Copilot MCP settings and the tools will be available in your coding sessions.
ChatGPT requires enabling Developer mode in Settings, then creating a new app with the provided MCP Server URL. Set authentication to noauth and your tools will be available in ChatGPT conversations.
Best Practices
Keep Skillsets Focused: Each integration exposes one skillset. Create purpose-specific skillsets rather than putting every ability into a single one. This makes it easier for AI clients to pick the right tools.
Protect Your Access Token: The access token grants full access to call all tools in the integration. Store it in a secure vault like 1Password or Bitwarden and use environment variables instead of hardcoding it in configuration files.
Name Abilities Clearly: MCP clients present tool names and descriptions to the AI model. Descriptive names and clear descriptions help the model choose the right tool at the right time.
Test Before Sharing: After setting up the integration, try each tool from your MCP client to make sure it works as expected. Check for timeouts, missing parameters, and unexpected responses.
Practical Use Cases
Developer Tooling: Expose internal APIs, deployment scripts, or database queries as MCP tools so developers can trigger them directly from their coding environment.
Knowledge Retrieval: Attach a skillset with document search abilities so any MCP client can look up company knowledge, product documentation, or support articles during a conversation.
Workflow Automation: Combine abilities like calendar access, email sending, and CRM lookups into a single skillset. AI assistants can then orchestrate multi-step workflows by calling the right tools in sequence.
Customer Support Agents: Give AI agents access to tools for checking order status, issuing refunds, or updating account details - all through a secure MCP endpoint.
Content Generation Pipelines: Expose tools for image generation, translation, or content formatting so AI applications can produce rich, multi-format output in a single session.