Skip to content

frankxai/starlight-academy

v0.1.0Apache-2.0

Practice with Academy teaching guides, source-backed challenges and useful AI feedback in your own conversation.

Starlight Academy

Public teaching material for your AI conversation: choose a challenge, make your own first attempt, practice with a guide, critique, revise and test transfer.

The plugin provides one guiding skill and connects four public MCP tools: get_academy, get_teacher, get_challenge, get_feedback_guide. The server is https://starlightintelligence.academy/mcp (Streamable HTTP). It accepts catalog IDs only. No drafts, accounts, private data, model calls, student database, automatic training, credentials or external actions.

Your chosen host provides the AI conversation under its own data policy. Teachers are fictional teaching methods, not separately running models or human instructors. Switching teacher roles is not independent review.

Connect and practice

Use https://starlightintelligence.academy/academy/connect for the current connection instructions and a matching browser challenge. The root plugin.json and mcp.json use Agent Plugins 1.0.0. Codex and Claude compatibility manifests are included. Host support varies; a standard manifest is not proof of execution in every host.

For Codex CLI, connect the remote tools directly:

codex mcp add starlight-academy --url https://starlightintelligence.academy/mcp

For Claude Code:

claude mcp add --transport http starlight-academy https://starlightintelligence.academy/mcp

The commands connect MCP tools. To install the bundled teaching skill as well, use a compatible plugin host with this complete folder or the public directory after it is published. A direct MCP connection is not a plugin-directory install.

Start a fresh conversation: “Use Starlight Academy. Help me choose a challenge and ask for my own first attempt before showing a solution.”

Publication status

This package is independently developed by Starlight Intelligence Systems. OpenAI directory submission is being prepared. It has not been submitted, approved or publicly listed; there is no OpenAI Verified claim. The existing offline starlight-department-lab package remains a separate five-skill toolkit.

Privacy and feedback

Read https://starlightintelligence.academy/academy/connect and https://starlightintelligence.academy/privacy. The server returns public material and processes selected catalog IDs. Hosting may process technical request data. The plugin does not collect feedback text or opt you into model training. You choose what to preserve or share through your host's tools.