Energy-matched blocks
You have peak hours and slump hours. Hard, deep work goes in the peak. Light tasks fill the slump. Fighting your own rhythm is the most common way to waste a study session.
For students · Free
Anchor — The science-backed, AI-managed calendar layer for students.
Anchor helps students maximize goal achievement and cognitive capacity using automated behavioral scheduling, completely free.
Your calendar should work with your brain, not against it.
Three steps. You stay in charge.
01 Dump it
Homework, exams, projects, the thing you keep avoiding. Drop it all in one inbox with a deadline and a rough size. No planning yet.
02 It plans
Anchor reads your calendar and finds the real free time. Then it fits every task into the slot where you will actually do it well.
03 You approve
Anchor proposes, you decide. Nothing on your calendar moves silently, and every block comes with a one-line reason.
Built on how people actually behave.
A schedule that ignores energy, memory and avoidance is a wish list. Anchor schedules around all three.
Spaced repetition
Review sessions are scheduled on expanding intervals (1 day, 3 days, then longer), and shorten again when recall is weak. Memory research says this beats cramming.
Avoidance, noticed
When the same task keeps getting skipped, Anchor notices and brings it back into view instead of letting it quietly rot at the bottom of a list.
Commitment risk
Every deadline is a promise. Anchor scores each one green, amber or red, so the promise you are about to break is visible days before, not the night before.
The algorithm does the math. The AI does the thinking.
Language models are bad at time arithmetic and good at reasoning. So Anchor splits the work, and keeps the human on the final decision.
Live today
- Deterministic scheduler. Free slots, task scoring and slot fitting run as plain code, so nothing overlaps and no deadline math is guessed.
- AI reasoning layer. A language model reads the finished plan and writes the why: the morning briefing, one reason per block, and a warning when the day is overloaded.
- Google Calendar, read-only. Anchor sees your real commitments. It never edits your calendar without your approval.
- Private by default. Every row of student data is locked to its owner with Postgres row-level security.
In development: built on Claude
- Multi-step reasoning with Claude. Turn an unstructured school timetable or a syllabus PDF into dynamic, balance-optimized time blocks: read it, extract the deadlines and topics, weigh them against your energy and workload, then lay them out across the weeks.
- Agentic calendar actions over MCP. Use the Model Context Protocol so Claude can act on a calendar through narrowly scoped, permissioned tools, with every change shown to the student for approval first.
- Same rule as today. The AI proposes. The student approves. Nothing moves silently.
What I'm building with Claude.
One product, one builder, and an honest account of where Claude sits in it today and where it goes next.
What it does
Anchor is a calendar layer for students. You drop in tasks, deadlines and exams. It reads your real Google Calendar, finds the free time, and proposes a realistic daily plan matched to your energy, your deadlines and your memory, with a one-line reason for every block. You approve it. Nothing moves silently.
Who uses it
Students juggling classes, exams and deadlines, especially in demanding programmes like the IB. I'm an IB student in Singapore building the planner I wish I had. It is early stage, and the app is live.
How Claude is part of it
Claude is how Anchor gets built today, and the model I'm moving its reasoning onto. Here is exactly what is live and what is not.
Live
Claude Code builds Anchor
The product design, database schema, scheduling logic and Next.js app are written and reviewed with Claude Code. It is how one student ships a full-stack product.
Next
Claude API as the reasoning layer
Replaces the language model that writes today's briefings and per-block reasons, and adds the hard part: multi-step reasoning over a messy school timetable or syllabus PDF, extracting deadlines and topics and laying them out as balance-optimized time blocks.
Next
MCP for calendar actions
Claude acts on the calendar through narrowly scoped Model Context Protocol tools, and every change is shown to the student for approval before it is made.
Why Teams
A workspace to build and test it
The Claude API is what runs inside the app. Claude for Teams is the shared workspace where I design the prompts, test them against real student schedules, and ship the code. Anchor is free for students, so the credits and seats are what make building it properly possible.