Knowledge.Dev can now work as an MCP server for your AI assistant. That means you can keep learning inside your normal coding workflow while the assistant uses the course state, progress tracking, and tutoring rules provided by Ride.

Install the Launcher

Install the knowledge launcher from Cargo:

cargo install knowledge

Then download or refresh the current Ride binary:

knowledge update

The launcher keeps Ride up to date and starts it when needed.

Connect It to Claude Code

The recommended setup for Claude Code is:

claude mcp add knowledge -- knowledge mcp

This registers knowledge as the MCP command and starts Ride in MCP mode behind the scenes.

Why this matters:

  • knowledge handles updates for you
  • Ride runs as a tutoring MCP server
  • logs stay out of the MCP stdio channel

Connect It to Codex

The simplest way to connect it to Codex is:

codex mcp add knowledge -- knowledge mcp

Start a Lesson

Once MCP is connected, ask your assistant to open a playground, for example:

playground rust-actor-framework

Ride is designed to behave like a teacher, not an autonomous implementation bot. At the start of a session the assistant should first ask what language you want to speak and which mode you want to use: Pure for direct engineering or Vibe for agent engineering. After that it can load the current step and guide you inside your workspace.

Ride is tuned for a strict tutoring flow. It should explain the goal of the current step, break the work into small actions, review your result with explanations, and update progress as tasks are completed. It should not write the solution into your project, silently complete the step for you, or patch files on your behalf. The point is simple: Ride should teach, check, and guide, while you still do the implementation.

When You Need More Help

If you get stuck, you can explicitly ask for a hint or for the reference implementation of the current task. That way you can keep learning step by step most of the time, and only peek at the answer when you really need it.

Good feedback should still stay educational. Instead of only saying that something is missing, the assistant should explain what is wrong, why it matters in the current step, and what to fix next.

Ready to try it? Install the launcher with cargo install knowledge, connect it as MCP, and start with the Rust Actor Framework playground.