ℹ️ Οι περιγραφές ρόλων των πρακτόρων είναι προς το παρόν μόνο στα Αγγλικά. Το περιβάλλον είναι στη γλώσσα σας.
BUILT WITH 15 AI AGENTS
One conversation.
Fifteen specialists.
CodeMentor AI is built by a team of 15 specialized AI agents working in parallel — project managers, tech leads, lesson authors, an 18-language translator, QA, pedagogy curators, and three user-persona testers. The result: 1,400+ lessons across 13 tracks, all translated, all reviewed.
🧠 Leadership
Talks to you. Plans. Coordinates everything.
#1lead-pm
Project Manager — your single point of contact
The only agent that talks to the founder. Translates directives into briefs, reports status with short yes/no questions, and never delegates the user-facing conversation downstream.
Read · Write · Edit · Bash
#2co-planner
Tech Lead — turns briefs into parallel work
Receives briefs from lead-pm, designs the technical plan, dispatches workers, and is the only one allowed to commit + push. Optimizes for parallel execution across doers.
Read · Write · Edit · Bash · Grep · Glob
🚀 Workers
Execute. In parallel. Don't debate scope.
#3doer-frontend
Next.js · React · TypeScript · Tailwind · i18n
Implements UI tasks dispatched by co-planner. UI components, pages, accessibility, i18n strings — never discusses scope.
Read · Edit · Write · Bash · Grep · Glob
#4doer-backend
FastAPI · Supabase · Postgres · API routes
Builds API routes, RLS policies, webhooks, DB migrations. Hard security checklist on every endpoint. No premature optimization.
Read · Edit · Write · Bash · Grep · Glob
#5lesson-author
Authors lesson batches in track-*.ts files
Reads existing track schema, appends N new lessons matching the file's exact style (READ+QUIZ vs Pyodide-runnable). Verifies with tsc before reporting done.
Read · Edit · Write · Bash · Grep · Glob
#6creative-senior-dev
Invents lesson formats that don't exist elsewhere
Code Detective. Time-Travel Refactor. AI Pair-Programmer-Reversed. Creates differentiators against LeetCode, Codecademy, freeCodeCamp.
Read · Edit · Write · Bash · Grep · Glob · WebSearch
🌍 Translator
Auto-fires after any content agent. 18 languages.
#7translator
Google free-endpoint chunked batch translator
Translates new lessons into 18 languages via translate.googleapis.com. Chunked for large content, jittered to avoid rate limits, runs unattended in the background.
Read · Write · Edit · Bash · Grep · Glob
🔍 Quality gate
Three filters every lesson must pass.
#8pedagogy-curator
Did the learner actually learn?
Checks difficulty curve, prerequisite chain, whether the lesson has an 'aha' moment, whether theory matches task. Blocks dry/boring/disconnected lessons.
Read · Bash · Grep · Glob · WebFetch
#9qa-controller
Build + i18n + security regression check
Runs tsc, build, and i18n coverage check. Verifies every language sees its own translation (no English fallback for RU/UK users).
Read · Bash · Grep · Glob · WebFetch
👥 User personas
Three AI-simulated personas. Three perspectives. Real friction surfaced.
#10persona-beginner
15-year-old ESL student, mobile-first
Plays through new lessons as a real teenage beginner whose English is a second language. Flags confusing jargon, dense theory, missing scaffolding.
Read · Grep · Glob · WebSearch
#11persona-career-switcher
32-year-old accountant, learning after work
Tired, motivated, time-boxed. Reports lessons that demand too much focus, demand context she doesn't have, or pace too fast for a 25-min evening study slot.
Read · Grep · Glob · WebSearch
#12persona-mid-dev
28-year-old Python dev, 3 years experience
Scans, doesn't read. Looks for depth. Defends L3-L5 tracks from content that's too shallow. Filters out things every senior dev already knows.
Read · Grep · Glob · WebSearch
💼 Business
Work on a cron schedule — distribution, legal, design.
#13growth-marketer
SEO · landing copy · Reddit · email · A/B tests
Owns distribution. Runs weekly. Tunes meta tags, writes hreflang-aware landing copy, suggests subreddits / HN / Twitter posts, plans 7-day drip emails.
Read · Edit · Write · Bash · WebFetch · WebSearch
#14legal-compliance
FTC · GDPR · CCPA · COPPA · PCI
Reviews every money flow BEFORE it ships. Critical for trial-then-auto-charge (FTC negative-option). Audits cookie consent, refund text, age gates, billing flows.
Read · Edit · Write · Bash · WebFetch · WebSearch
#15product-designer
UI / UX review and polish
Reviews screens, suggests interaction improvements, owns the visual polish on every new UI feature. Works through doer-frontend for actual edits.
Read · Edit · Write · Bash · WebFetch · WebSearch
What happens when you ask for 60 lessons
One request flows through 8 stages. Most run in parallel; total wall time is hours, not weeks.
1
You write the founder
‘Add 20 lessons to the FastAPI track’
2
lead-pm receives it
Translates RU → brief: track / range / topics
3
co-planner plans
Splits into 3 parallel sub-tasks
4
3× lesson-author run in parallel
20 lessons each, ~1 hour total wall-time
5
translator fires automatically
18 languages, 3000+ translation jobs in the background
6
pedagogy + qa + 3 personas
Three independent quality filters
7
co-planner pushes to main
Vercel auto-deploys
8
lead-pm reports back
‘60 new lessons live, 18 languages, build green — anything else?’
Without agents vs with agents
WITHOUT AGENTS
- ✗1 lesson ≈ 30 min by hand
- ✗You translate manually, one language at a time
- ✗Single tester, single perspective
- ✗Design / legal / compliance — "later"
- ✗Content velocity caps out at ~5 lessons / day
WITH AGENTS
- ✓60 lessons in 1 hour — three parallel authors
- ✓18 languages translated automatically in the background
- ✓3 personas + pedagogy + QA — every lesson reviewed 5 ways
- ✓Cron-scheduled growth, legal, design — never "later"
- ✓Content velocity ≈ 60+ lessons / day, sustainably
Try what the team built
1,400+ lessons. 13 tracks. From Python basics to senior-level interview prep, FastAPI, AI engineering, and data science — in 18 languages.
Browse all 10 tracks →