LearnCo is the cognitive layer for EdTech. Stream learner activity in; get back a real-time, principled read of what each learner understands, where they're confused, and what to teach next. The instrumentation your product is missing — not another dashboard on top of the one you already have.
import { LearnCo } from '@learnco/sdk' const lc = new LearnCo({ apiKey: process.env.LEARNCO_KEY }) // Capture — stream a learner event in await lc.capture.event({ learnerId: 'lrn_a8f3', concept: 'statistics.hypothesis_testing', result: 'incorrect', duration: 42, }) // Decisions — ask what to teach next const next = await lc.decisions.next({ learnerId: 'lrn_a8f3' }) // → { concept: 'statistics.sampling_distributions', reason: 'prerequisite gap' } // Insights — query the learner model const model = await lc.insights.learnerModel({ learnerId: 'lrn_a8f3' }) // → { mastery: 0.34, gaps: ['sampling_distributions'], readyFor: ['z_scores'] }
Engagement is not a proxy for learning. It is a different measurement entirely. Quiz-pass rates measure performance under test conditions. Streaks measure habit maintenance. Completion measures clicking a button. None of them tell you what the learner actually understands — and every personalization system built on them inherits that gap.
This is not your team's fault. The tooling does not exist. Mixpanel measures events. Amplitude measures funnels. Nothing on the market measures learning. LearnCo is the attempt to close that gap.
LearnCo is an API platform that adds one capability to your product: a real-time, principled read of what each learner has actually learned — not what they clicked, completed, or guessed right.
You stream learner activity in. We stream decisions back — what they've mastered, what they're getting wrong and why, what to surface next, when to intervene.
If your engineers know Amplitude, they'll know LearnCo's shape on day one.
Same shape your team already knows from Mixpanel, Amplitude, or PostHog — with a learning-aware schema underneath.
await lc.capture.event({ learnerId: 'lrn_a8f3', concept: 'stats.hypothesis_testing', result: 'incorrect', duration: 42, meta: { attemptNumber: 3, hintUsed: true }, }) // → { eventId: 'evt_9k2m', processed: true }
Mastery updates. Misconception flags. Intervention recommendations. Bidirectional stream, not request/response.
const next = await lc.decisions.next({ learnerId: 'lrn_a8f3', }) // { // concept: 'stats.sampling_distributions', // reason: 'prerequisite gap', // confidence: 0.87, // }
In aggregate or individually. The dashboard your curriculum team has been quietly building in spreadsheets.
const model = await lc.insights.learnerModel({ learnerId: 'lrn_a8f3', }) // { // mastery: 0.34, // gaps: ['sampling_distributions'], // readyFor: ['z_scores'], // risk: 0.71, // }
The substrate underneath all three surfaces is the Human Knowledge Index — a structured knowledge graph that maps what humans need to learn and in what order, and the taxonomy of how learners misunderstand — the bugs, p-prims, and partial models behind real mistakes.
Longitudinal mastery signal flows in from real learners, across every integration. Every EdTech that plugs in makes the index sharper. Every learner sharpens it for every other product that uses it.
Who know engagement metrics aren't telling them the truth.
Who want personalization that actually works — not adaptive UX dressed up as adaptive learning.
Who'd rather drop in an SDK than build a learner-modeling stack from scratch.
Who want their domain expertise reflected in the runtime, not stuck in a PDF.
If you're shipping any product where someone is supposed to learn something — test prep, professional skills, K-12, corporate L&D, language, coding, AI tutors — this is for you.
We made them on behalf of the learner. Every EdTech that integrates keeps what it built; LearnCo adds what it couldn't build alone. If any of these constraints ever stop serving learners, we'll say so publicly and reconsider.
The learner identity is minted by LearnCo and held by the learner. Your product binds it to your internal user-id. The learner can revoke at any time.
Your curriculum is yours. We don't write lessons, items, or copy.
We're not building a learner-facing app. Your product owns the relationship.
The whole point is that no one product owns the learner. Including us.
Someone trying to understand a concept they've struggled with for months deserves a system that actually knows whether they understand it. That is the reason this exists — not the integration count, not the press, not the market size.
We're currently running a proof-of-concept on a narrow, well-mapped domain — to validate that the underlying diagnostics actually work before we generalize. The research artifacts, design decisions, and roadmap are written in the open.
If the thesis is right, LearnCo becomes infrastructure. If it's wrong, the research gets published and the next team builds on it. Either way, we are willing to be wrong about it in public.
Be among the first to integrate when the SDK ships. We'll only email when there's something real to share — no growth-hacker drip campaigns.
We'll reach out when there's something concrete to integrate against. In the meantime, occasional notes on what we're learning.