knot knot

AI

Starting with version 0.19.0, Knot introduces AI support through two key features:

  1. Model Context Protocol (MCP) When MCP functionality is enabled, the Knot server provides tools that can be utilized by compatible clients. These clients may include editors like Visual Studio Code or desktop applications such as ChatWise, provided they support tool calling and the OpenAI API.

  2. Web-Based Assistant Enabling the web-based assistant allows users to interact with their spaces and their contents via a chat interface. This assistant requires the Knot server to be connected to a Large Language Model (LLM).

    Additionally, when the web-based assistant is enabled, clients that do not support tool calling can connect directly to the Knot server. In this setup, tool calls are handled internally by the server, leveraging knot’s internal system prompt.

What’s Next

  • Slash Commands: Create and use slash commands as reusable prompt templates in the AI assistant chat.
  • MCP: Enable and configure knot's Model Context Protocol server and connect MCP clients.
  • MCP Tools: Built-in MCP tools for managing spaces, templates, stacks, files, and commands in knot.
  • Remote MCP: Connect external MCP servers to knot for its own AI features (web chat, OpenAI-compatible API, scripts).
  • AI Assistant: Configure knot's web-based AI assistant and CLI, which connect to an LLM provider such as OpenAI, Anthropic Claude, Google Gemini, Ollama, Mistral, Z.AI or xAI Grok.
  • Skills: Author and manage markdown skills that give AI assistants context and procedures in knot.
  • System Prompt: Customize the system prompt the AI assistant uses to operate within the knot environment.