MindTrain plugin
A Codex plugin for daily knowledge training: the AI asks questions, your own server scores answers and schedules reviews.
- 129 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The plugin runs a local Python MCP bridge and reads the materials folder you point it to.
- The `MINDTRAIN_BOOTSTRAP_TOKEN` token opens access to your instance, so keep it out of the repository.
Install
- Works in
Manual install
git clone https://github.com/shigella520/MindTrain.gitThe server side. Then, in `deploy/core-only`, copy `.env.example` to `.env`, set the PostgreSQL password and token, and run `docker compose up -d --build`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The plugin runs a local Python MCP bridge and reads the materials folder you point it to.
- The `MINDTRAIN_BOOTSTRAP_TOKEN` token opens access to your instance, so keep it out of the repository.
This is third-party code. Review the repository files before installing.
What it does
MindTrain turns what you want to learn into daily questions. The Codex plugin adds the `$mindtrain` skill and a bridge to Trainer MCP. The AI asks questions and explains, while a Java and PostgreSQL server checks answers, records history and chooses what to review. You can argue with any question, ask for a hint or find out why an option is wrong: only an explicitly submitted choice counts as an answer. The single source of data is Training Core on your own server. It limits new questions per day and pauses them when the review backlog grows. Topics are built through dialogue or from your folder of notes in MD, TXT, PDF, DOCX and PPTX.
Who it is for. People preparing for an exam or interview who want to practice in dialogue with Codex on their own server instead of someone else's cloud.
Good fit when
- You prepare for a Kubernetes interview and want daily practice on Pods, Deployments and Services.
- Notes in PDF and DOCX should become a topic tree and quiz questions.
- An answer is graded strictly by the selected options, and the model's opinion does not change the score.
- Old questions should return on a schedule so forgotten material does not pile up.
Not a fit when
- You have no Docker or server of your own: the plugin does not work without Training Core.
- You need shared training for a team: MindTrain is built for a single user.
- You need task types other than single and multiple choice.
Example request
$mindtrain Start a Kubernetes training session and after each answer explain why the other options are wrong.Limitations
Docker with Docker Compose, four containers (Core, Trainer MCP, Web, PostgreSQL) and your own `MINDTRAIN_BOOTSTRAP_TOKEN` are required. The plugin targets Codex, and no other clients are described in the README. PDF parsing extracts text only, with no scan recognition. The Anki and FSRS scheduler is still planned, and a weighted scheduler runs now. The project is built for a single user. Back up PostgreSQL before upgrading: Flyway migrates the schema at startup, and a failed migration is rolled back from the dump.
How to disable. Stop the server with `docker compose down` in the `deploy/core-only` folder, and the data stays. Remove the plugin through the Codex plugin manager; the settings live outside the project.
FAQ
How do I install the MindTrain plugin in Codex?
Run `codex plugin marketplace add shigella520/MindTrain --ref main` and `codex plugin add mindtrain@mindtrain`. Then, in a new task, enter `$mindtrain Configure my MindTrain instance` and give the MCP address and token.
What does the server store?
Training Core keeps questions, sessions, answers, review state and settings in your PostgreSQL. The skill stores nothing, and the original material files stay on your machine.
Does a question stay open if I ask for clarification?
Yes. An answer counts only after an option is explicitly submitted, so clarifications and hints do not use up the question.
How do I upgrade MindTrain?
Take a PostgreSQL dump, run `git pull --ff-only` and `docker compose up -d --build`. Upgrade the plugin with `codex plugin marketplace upgrade mindtrain` and `codex plugin add mindtrain@mindtrain`, then start a new task.
Which question types are supported?
Single and multiple choice. Grading compares the set of selected options, so a model cannot assign a score by feel.
Discussion
No questions or reviews yet.
Related
PluginsA Codex and Claude Code plugin where a coach teaches on real projects, guides debugging and records evidence of mastery
Plugin, MCP server and app for a Markdown knowledge base: your AI tool maintains pages while you see changes and citations.
Claude Code plugin that builds a study plan for any topic, runs adaptive quizzes and schedules reviews with the SM-2 algorithm.
A Claude Code plugin that builds a knowledge system of markdown files from a conversation with you: folder structure, note templates, a processing pipeline and
Study coach plugin for people with ADHD: small chunks of new material, review quizzes and explain-it-back prompts.
An agent writes questions and exams on EduBase, assembles quizzes and analyzes results, using an integration key
README in short
The project consists of Training Core, Trainer MCP, a web interface, a Codex plugin and a weighted scheduler. The web page shows a dashboard, a training map and question bank management. Plugin and server versions are checked automatically, and incompatible versions stop training.