Faroz

Faroz MCP server — remote MCP for a kids' education platform

Endpoint: https://faroz.uz/mcp · Docs: faroz.uz/faroz-mcp · Registry: uz.faroz/faroz · Auth: OAuth 2.0 (no API key)

Connect AI assistants (Claude, ChatGPT, Gemini and any MCP client) to Faroz, a learning platform for children aged 8–17. Teachers, parents and content authors use it to plan lessons, create learning content, run the lesson calendar, check students' homework and build age-appropriate habits by simply talking to the assistant.

Transport: Streamable HTTP. Auth: OAuth 2.0 with dynamic client registration — the client opens the Faroz sign-in page on first use.

What you can do

Exercise block types (11 + 2 layouts)

Text multiple choice · picture multiple choice · fill in the blanks · typed answer · speak-aloud answer (speech recognition) · image · image gallery · audio · video · styled rich text · callout boxes (rules, examples, tips)

Layout blocks: column (blocks one under another — e.g. several blocks inside one callout) and row (2–3 blocks side by side with equal widths, on phones too).

AI-graded pages

• Grammar check (grammar_check) — the student photographs written English work; AI lists the grammar mistakes and the CEFR level. Mixes with block pages.
• Speaking exam part (language_speaking) — one CEFR Multilevel or IELTS speaking part; the student answers by voice and AI grades the recordings. One part per exercise.

See get_exercise_format_guide(topics=["page_types"]) and ["language_speaking"].

Tools

Read-only: whoami, search, fetch, list_courses, get_course, get_course_filters, get_course_schedule, list_course_students, list_homework_submissions, get_homework_submission, list_exercises, get_exercise, get_exercise_usages, get_exercise_format_guide, validate_exercise_body, list_game_categories, list_games, list_media, list_subjects

Create: create_course, add_lesson, create_exercise, create_game, add_calendar_lesson, create_recommendation, create_strategy, add_strategy_recommendations

Edit (overwrites existing content): update_course, update_lesson, reorder_lessons, reschedule_lesson, set_group_schedule, update_exercise, update_game, update_media, review_homework, update_recommendation, update_strategy, reorder_strategy_stages

There are no delete tools and no payments: prices, subscriptions and wallets are never shown or changed through this server. Every call runs as the signed-in user, with exactly the same permissions as in the Faroz app. Homework tools work only for the course's staff (teacher, support teacher, course admin).

Building a complete course (built-in playbook)

Every client gets the same course-authoring playbook automatically — no setup needed:

• Server instructions tell the assistant to read it before creating a course.
• Tool guide: get_exercise_format_guide(topics=["course_authoring"]) returns the full playbook (also the resource faroz://docs/exercise/course_authoring).
• Prompt: create_course (arguments: topic, optional audience, lessons) — in Claude pick it from the prompt menu, in Gemini CLI run /create_course.
• Tool results: create_course returns next_steps; create_exercise / update_exercise return validation_warnings (e.g. a test without a question, unstyled text) so the assistant fixes them.
• Skill file: skills/faroz-course-builder/SKILL.md — upload it as a skill in Claude, or use it from the ChatGPT/Codex plugin.

What the playbook enforces: full course metadata and a banner, 4–6 lessons with 2–3 exercises each (theory → practice → test, final test at the end), all 11 content block types across the course (plus column / row layouts where they help), a question above every answer block, styled text and callouts, verified image/video/audio links, homework on 1–2 lessons, building with is_active=false and publishing only after a final check.

Homework

Students submit homework for a lesson in the Faroz app: images, audio, PDF, Word or presentations, up to 20 MB in total, plus an optional comment. Files are stored privately and served only through signed links that expire. Teachers review them in the app or with an AI assistant through this server.

Install

Claude: Settings → Connectors → *Add custom connector* → URL https://faroz.uz/mcp → Connect → sign in with your Faroz account.

ChatGPT: Settings → Apps & Connectors → Advanced → Developer mode → *Create* → MCP Server URL https://faroz.uz/mcp, Authentication: OAuth.

Claude Code

claude mcp add --transport http faroz https://faroz.uz/mcp

Cursor, VS Code, Windsurf and other `mcp.json` clients

{

"mcpServers": { "faroz": { "url": "https://faroz.uz/mcp" } } }

Gemini CLI

gemini extensions install https://github.com/Akramjon-koder/faroz-mcp

Any MCP client: add a remote server with the URL https://faroz.uz/mcp. The client registers itself (RFC 7591), then opens the Faroz sign-in page (email/password or Google).

Example prompts

• "Create a 5th-grade math course on fractions with 4 lessons and 2 exercises in each lesson."
• "Make a beautiful course about the Solar System for 9–13 year olds." (the assistant follows the course playbook: banner, video, pictures, tests, homework)
• "Here is a paragraph about the water cycle. Make a multiple-choice, a fill-in-the-blank and a typed-answer exercise from it, then add them to lesson 2."
• "Open my fractions course, check every lesson's exercises for mistakes or unclear questions, and fix what you find."
• "Show my courses and list the exercises in the first lesson."
• "Check all homework waiting for review in my fractions course, grade each one and write short feedback for every student."
• "Who are the top 5 students in my course by mastery?"
• "Move tomorrow's lesson of my fractions course to Friday at 16:00."
• "Build a 'beating laziness' strategy for 10–12 year olds from fitting recommendations and add one new daily recommendation for November."

Requirements

A free Faroz account: sign up at faroz.uz. No API key is needed.

Roadmap

Possible next steps:

• Per-student progress reports for teachers and parents
• Help for students with their lessons and learning progress

License

The files in this repository (extension manifest, context file and skills) are released under the MIT License. The Faroz platform and the MCP server behind https://faroz.uz/mcp are proprietary services.

Privacy & support

• Privacy policy: https://faroz.uz/privacy
• Terms: https://faroz.uz/terms
• Support: akramjonminochdinov@gmail.com