Hover any move — it plays on the board. Click a question to take over.
400
games read in ~60s
100%
of claims engine-verified
0
made-up moves. Structurally.
1100–1900
human play, modeled per rating
only 6% of 1400s find Nxf7you score 66% with queens on · 45% off earlyblunders cluster in 15–60s thinks at 2.3×leaves book at move 3.4 on averagemove 22: mate in 2 — missedone mistake breeds another at 4.6×24% of your level find Na555% of mistakes die to a forcing moveonly 6% of 1400s find Nxf7you score 66% with queens on · 45% off earlyblunders cluster in 15–60s thinks at 2.3×leaves book at move 3.4 on averagemove 22: mate in 2 — missedone mistake breeds another at 4.6×24% of your level find Na555% of mistakes die to a forcing move
Everything is measured
One coach, every angle.
It read your games. All of them.
I’ve been through your games. Positional inaccuracies — 57× across 55 games. Blunders cluster in 15–60s thinks (2.3×). One mistake breeds another at 4.6×.
A real first message — every number measured from your games, none generated.
Your blunder fingerprint.
Blunder rate by think time
<5s
0.7×
5–15s
2.1×
15–60s
2.3×
60s+
0×
When you blunder, what you misjudge, where your prep ends.
Human context.
6%of 1400s find Nxf7 24%find the right Na5 — on every single move
Not just engine-best — what people at your level actually find.
Spar The Human.
The Human (1400) real openings · real plans real occasional slips
A bot that plays like your next opponent, not like a computer.
Puzzles that fit.
fork · rated 1340 9% of your level solve it
Aimed at your recurring mistakes, graded by real human solve-rates.
Every game reviewed.
Eval — you vs. opponent
Each turning point found and explained in plain language.
Zero hallucinations.
you ask→engine computes→verified→coach speaks ✓
If a claim can’t be verified, it isn’t reworded. It’s removed.
One coach, every question
Ask it anything.
ReviewWhy did I lose this game?
ReviewWas Nxe5 a blunder?
LearnTeach me the London System
LearnWhat should I play against 1.e4?
LearnHow do I win K+Q vs K?
AskIs this position winning?
InsightsWhat's my worst opening?
InsightsWhat mistakes do I repeat?
TrainPlay against me like a 1600
TrainGive me a puzzle from my weaknesses
One minute to start
How it works.
1
Connect
Drop in your Chess.com or Lichess username. No password needed — games are public.
~10 seconds
2
The coach reads
Your games get scanned move by move. Recurring mistakes, habits, openings — measured.
~60 seconds
3
Train what it found
The first message is your personal brief. From there: explanations, puzzles, drills, sparring.
the rest is chess
Join 100+ players sharpening their game
Early players, real gains.
★★★★★
“First app that told me WHY my rook was bad instead of just slapping a ?? on it. Up 80 points in three weeks.”
Jordan M.1450 rapid
★★★★★
“The Human bot is unfair — it plays exactly like the 1600s I keep choking against. Finally practicing the real thing.”
Priya K.1580 blitz
★★★★★
“Asked why I keep losing with the Caro and it just… knew. Pulled up my actual games. Kinda spooky honestly.”
Devin R.1320 rapid
★★★★★
“I don’t trust AI for chess. This one literally cannot make moves up — that’s the whole point. Sold.”
−20% Launch price — locked in while you stay subscribed.
✓Every game deep-scanned with high-depth Stockfish
✓200 coach questions a month
✓Unlimited puzzles, drills & sparring
✓Full tendencies profile & progress tracking
Get PremiumCancel anytime from your profile, one click.
Questions players actually ask
FAQ.
Does it make things up like other AI chess tools?
No — by construction. A real engine (Stockfish 17) computes the analysis first; the coach only narrates what was computed, and every move it mentions is verified before you see it. If a claim can’t be verified, it’s removed. That’s the whole architecture.
Which platforms does it work with?
Chess.com and Lichess — just your username, no password (your games are public). You can also paste any single game link or a position (FEN).
How many games does it need to know me?
The coach is useful immediately with a single game. The personal profile — recurring mistakes, think-time habits, queen-trade record — gets solid around 50+ analyzed games and reads your last 400.
What level is it for?
Built for roughly 800–2200. The human-play model covers 1100–1900 in detail, so “what do people at my level do?” has a real answer — not a guess.
What does free actually include?
Importing and analyzing your games, coach chat with monthly credits, daily puzzles, and the bots. Premium ($8/mo, or $70/yr) raises the limits to 200 coach questions a month and deep-scans every game you’ve imported.
Can I cancel Premium?
Yes — one click from your profile. It stays active through the end of your billing period, no surprise charges.
From the founder
Hey — I’m Matt. I’m 16 and still in high school, and I built Deepline because I was sick of chess apps telling me what I did wrong but never why. A “??” with no explanation doesn’t teach you anything.
So I made the coach I actually wanted: one that reads all my games, remembers the mistakes I repeat, and explains them in plain words — grounded in a real engine so it can’t bluff. It’s just me building this, and every piece of feedback comes straight to me.
MMatt16 · founder & solo builder
Your games are the curriculum.
Connect Chess.com or Lichess. Meet your coach in a minute.