Open source · MIT licensed
Your next chapter starts here.
Get Tutor ACP to learn with your community. Get Tutor MCP to run your own learning engine. Both are open source; your AI provider account stays separate.
Verified release links · live updates when GitHub is availableTutor MCP v0.7.0 ↗
Your desktop learning space
Tutor ACP0.0.1
Your courses, conversations and progress. One place to learn with Codex or Claude and join your institution’s catalogue.
Windows x64 · 337 MB · Codex and Claude adapters included
The desktop installer is currently available for Windows x64. On macOS or Linux, connect another MCP client to the Tutor MCP server.
Client release notes ↗- Install the app.No separate Node.js or Codex Desktop installation.
- Connect your agent.Choose Codex or Claude and sign in with your provider.
- Choose where you learn.Join your institution or install Tutor MCP locally.
Tutor MCP.
Run it where you choose.
Keep a local learning profile or host a shared service. Choose the server build for your computer. Tutor MCP is separate from the desktop app and the language model.
On a Mac, choose ARM64 for an Apple M-series chip or x64 for Intel. Windows ARM64 and Linux ARM64 downloads here are server builds.
tutor-mcp_linux_amd64.tar.gz13.0 MBOr install the latest stable server.
curl -fsSL https://tutor-mcp.dev/install.sh | shLinux and macOS · detects your processor, installs the latest stable server and verifies SHA-256.
Run tutor-mcp --version to check the installed version. Then connect your local client or deploy an institution.
Join your institution
Set up your identity, connect your agent and open a shared course.
On your computerStart locally
Connect the engine to Tutor ACP or an existing MCP client.
For operatorsHost a community
Plan PostgreSQL, HTTPS, keys, members and your first programme.
Tutor is MIT licensed and currently in alpha. AI provider accounts and any associated costs are separate. Installing open-source software does not include a hosted Tutor service.
No GitHub account is needed to download. If a release check is unavailable, the last verified links remain available. See the latest stable release on GitHub ↗.