Cert prep, language, upskilling.

The learning layer for AI

AI can explain anything. Sapior makes sure the learning sticks.

controlled studies

270+

Retrieval practice — pulling a fact out of memory builds memory — is one of the most replicated findings in learning research.

of memory science

50 yrs

From Ebbinghaus's forgetting curves to modern spacing research, Sapior is built on findings the field keeps confirming.

spaced-repetition engine

FSRS v6

The modern open-source algorithm modern spaced-repetition apps are built on. Tuned to schedule review at the moment your retention starts to decay.

Why generic AI breaks down for learning

Chat is a great tutor for one conversation. It forgets you the moment that conversation ends.

No memory between sessions

Every quiz starts from zero. The AI doesn't know what you got right yesterday or what's about to fade today.

Random review, not spaced

Generic chat picks topics by recency or curiosity, not by what you're statistically most likely to forget right now.

Plans drift, then disappear

A plan built in one chat doesn't carry into the next. Nothing tells you whether you're actually on track for your goal.

How Sapior works

A learning engine your AI can call

Sapior is an MCP server that gives any AI agent — Claude, ChatGPT, Cursor — durable memory for learning. State a goal. Sapior generates a plan, schedules what to study next, and tracks retention across every session, in every tool you use.

How it works

Sapior plugs into the AI assistant you already use. You bring the goal; Sapior runs the schedule.

Step 1

Set a goal

Tell Sapior what you're trying to learn. A certification, a language milestone, a body of knowledge. Sapior turns it into a plan you can adjust.

Step 2

Study with your AI

Sapior tells your AI assistant what to teach, review, or test next. You have a normal conversation; Sapior tracks what stuck and what didn't.

Step 3

Recall when it counts

Sapior schedules review at the moment retention is decaying, so the knowledge survives past the conversation. When you're in the exam or the meeting, you don't need the AI.

Why it works

Three findings learning research keeps validating

Decades of replication keep confirming what makes study stick. Sapior is built around the three findings that survived.

Retrieval is the learning event

Pulling a fact out of memory builds memory more than re-reading it does. It's one of the most replicated findings in learning research — the basis of more than 270 controlled studies. Every Sapior session asks you to recall, not re-read.

Spacing has a precise shape

How long you wait between reviews changes whether you remember. There's an optimal gap, and it shifts as you learn. Sapior schedules each review at the moment your retention is starting to decay — using FSRS v6, the engine that turns that science into math.

Scaffolding fades as you master it

Brand-new material gets taught. Once it's sticking, Sapior switches to guided review, then to pure recall. The supports that help when you're new actively get in the way once you've learned — so they shrink as your retention grows.

Who Sapior is for

Anyone with a learning goal that won't fit in a single session.

Cloud, security, finance, language certs

Certification candidates

Long study runways, dense material, fixed deadlines. Sapior keeps the plan honest as the date approaches and the surface area grows.

Upskilling and reskilling

Career switchers

Months of new domain material with no syllabus to anchor it. Sapior builds the structure, schedules the work, and tracks what stuck.

People who already learn with Claude or ChatGPT

AI-native learners

Keep the workflow you have. Add a memory layer that turns repeated chats into cumulative progress instead of restart loops.

Available where you already work

Sapior is MCP-first. Connect from any compatible client.

Claude Desktop

One-click install via the Claude Desktop Extensions directory.

ChatGPT

Listed in the ChatGPT App directory through the OpenAI Apps SDK.

Cursor

Compatible via Cursor's MCP support.

Any MCP client

Connect from any MCP-compatible host using https://sapior.ai/api/mcp

Frequently asked

If something's missing, ask your AI. It can call Sapior and answer.

What is MCP and why does it matter?

Model Context Protocol is the open standard for letting an AI assistant call external tools and remember state between sessions. Sapior is an MCP server, which means any compliant AI — Claude, ChatGPT, Cursor — can use it without integration work on your end.

Does Sapior generate the actual quiz questions?

When you study through Claude or ChatGPT, the host AI generates the question text from Sapior's directives — what concept to test, at what depth, given your retention. Sapior owns the curriculum and scoring; the AI handles presentation. The web wrapper uses Sapior's own LLM end-to-end.

What happens to my study history if I switch AIs?

Nothing. Your plan and retention live in Sapior, not in your AI's chat history. Connect a different AI tomorrow and pick up exactly where you stopped.

Can I bring my own course materials?

Small inline source context — pasted notes, excerpts, study guides — is supported today. Full-textbook ingestion with chunking and indexing is on the roadmap.

Is this a credentialing service?

No. The progress estimate is a rolling prediction built on two inputs: how much of your goal you've covered, and a model of your retention for each item you've studied. The retention model is FSRS v6 — the modern spaced-repetition engine, backed by decades of memory research. It's not a pass/fail score, and Sapior makes no outcome guarantees in any domain. See /methods for the science behind the engine.

What does it cost?

Free during the discovery phase, with a usage cap. Pricing comes after we know the engine works for you.

Soft launch, invite only

Add Sapior to your AI assistant in five minutes

One MCP connector, then ask your assistant to study with you. No app to install, no account to migrate.