Khanmigo

Khanmigo — Socratic AI Tutor Inside Khan Academy (Nonprofit-Backed, Active, Not Shut Down)

Not shut down.​ Khanmigo (khanmigo.ai / Khan Academy) launched Mar 2023, built on GPT-4/educationally fine-tuned model, nonprofit Khan Academy (Sal Khan), active 2026-08, no acquire/rebrand/shutdown. It’s the AI layer on top of Khan Academy’s 10,000+ lessons / 160M learners; the base platform stays 100% free, Khanmigo is the paid (or teacher-free) add-on. Category ai-productivity-data; pipeline_stage none. Killer differentiator = refuses to give direct answers — uses Socratic questioning to walk students to the answer (“productive friction”) + native tie-in to Khan Academy’s exercise graph so tutoring draws on the same videos/practice the student is already doing + under-13 safety guardrails + parent can see chat history + one $4/mo parent sub covers up to 10 children; not ChatGPT (gives answers, no curriculum tie-in, weaker K-12 guardrails), not Duolingo Max (language-only), not MagicSchool/Brisk (teacher-tool-first, no learner tutor at this depth), not Photomath/Symbolab (math-answer engine, opposite pedagogy). 2026 trap: only U.S. billing-address learners/parents can self-subscribe (international waitlist), only ~15% of eligible students actively use it​ (led to 2026 redesign, proactive help in assignments), teacher tools free in 44+ countries but teachers cannot grant students access — classroom use requires a Districts partnership (Clever roster, SSO, admin dash), Socratic refusal frustrates deadline-driven kids, subject depth uneven outside math/science/econ/CS, no FERPA/COPPA opt-out nuance (district data agreements only), model still hallucinates occasional wrong steps.

1. One-line positioning

Khanmigo = patient AI tutor, not homework-answer machine: sign in (parent account for under-18s) → pick a Khan Academy lesson or just ask → Khanmigo asks guiding questions, breaks the problem into steps, pulls the relevant KA video/practice → student works it out → writing coach critiques essay structure without rewriting it → Debate Me argues both sides → coding sandbox (Python/JS/Java + 9 more, auto-grading, 2026 Khanmigo 2.0) → SAT/ACT/AP prep with personalized practice → teacher side: standards-aligned lesson plans, rubrics, exit tickets, quiz questions + answer keys, differentiation, student-work summaries, class insights dashboard. Fits ai-productivity-data; pipeline_stage none (ask→guide→practice→assess→report). For K-12 learners inside the KA ecosystem, homeschool families, and U.S. teachers drowning in prep, it’s uniquely safe + cheap; for university/grad, non-KA curricula, or “just give me the answer,” it’s the wrong shape.

2. Core features

3. Pricing (2026-08, USD)

pricing_type = freemium-nonprofit-backed; starting_price = $0; free_tier = "yes; Khan Academy platform 100% free forever; U.S. teachers full Khanmigo tools free; learners/parents pay $4/mo or $44/yr (covers up to 10 kids)"; pricing_note = "khan-academy-nonprofit-ai-costs-covered-by-donations-subscriptions-not-student-data; teacher-free-via-microsoft-partnership-44-plus-countries-2026; parent-sub-4-mo-44-yr-up-to-10-children-included; district-access-requires-partnership-teachers-cannot-self-grant-students; us-billing-address-only-for-learner-self-subscribe; no-premium-tier-upsell-single-price; sales-tax-extra-on-4"

4. Access Type

access_type: web-app-plus-mobile-no-desktop

access_display: 🌐 khanmigo.ai + khanacademy.org (browser, account required) · 📱 Khan Academy iOS/Android (Khanmigo inside) · 🔌 Districts: Clever roster + SSO, Google Classroom/Schoology adjacent · no public API, no MCP, no browser extension, no offline

5. Reviews (2026)

Praise: Socratic method actually teaches vs ChatGPT answer-dump, $4/mo for 10 kids is unbeatable vs human tutoring ($30–80/hr), teacher tools genuinely save 45+ min/day (rubric + exit ticket in 15 min), curriculum-native means fewer wild hallucinations, safety guardrails + parent visibility best-in-class for under-13, nonprofit so no student-data monetization, writing coach critiques without rewriting (defensible for schools).

Gripes: only ~15% eligible-student activation​ (engagement, not quality problem), Socratic refusal infuriates kids up against a deadline, U.S.-billing-only locks out most of the world, teachers can’t just flip a switch for their class (need district partnership — slow), subject depth thin beyond math/science/econ/CS, advanced/AP Calculus+/grad-level unsupported, no API/MCP/extension (locked in KA ecosystem), model still makes occasional wrong steps, “next-item correctness” metric debated by learning-science researchers, humanities/language support weaker.

6. Best for / Not for

7. Competitors