CS2 AI coach
An AI coach that only talks about your match
Ask "why did we lose round 7?" or "could I have traded that?" and get an answer built from the demo: positions, timings, duels and utility of that exact round, with references you can click to see the moment in the viewer. Every answer is tied to what actually happened in your match.

How the coach answers
The coach does not guess from a summary. It queries the analysed match with investigation tools — the findings, the round reconstruction, the duels, the utility, the economy — and writes an answer that cites what it used. References look like [R7 · 22s] and jump to that second in the 2D viewer.
Before you read an answer, a fact checker verifies its round references, kills, grenades and player names against the demo, so what you read is backed by the match itself.
You choose the style: natural (how a coach would explain it), technical (shorter, denser, with the numbers) or brief.
Questions it is built for
- "Why did we lose round 7?" — the sequence of the round, the decisive duel or the lane that was lost, and the alternative.
- "Was that death tradeable?" — the nearest teammate, the distance, the seconds between the kills, and whether a trade was actually possible.
- "What should I fix first?" — the validated findings with the biggest impact, ordered, not a list of everything.
- "What did the opponent do better?" — their patterns across the match, with the rounds that show them.
- "Compare the two halves" or "what happened in the 5v3?" — any open question that the demo can answer.
What a grounded answer looks like
The coach works from the same findings the diagnosis shows. For the analysed team of a real FACEIT match on Mirage, these are the facts it reasons from:
Finding · R19 · 16 s
At 16 s the Ts took Mid after the CTs lost presence there at 15 s after losing a duel (Cac0_ died there).
A question like "why did Mid keep falling?" is answered with this round and the other five where a lane was lost before the site fell, not with a general remark about map control.
Evidence: Lane control finding, validated; rounds 13, 17, 18, 19, 21, 23.
Finding · R7
Kultera was left alone against 3 and won the round with 3 kills.
Asked "what went well?", the coach can point at this clutch and at the eco round won against a full buy in the same round, because both are recorded findings.
Evidence: Kultera wins a 1v3 with 3 kills.
Every reference opens the viewer

Answers are checked against the viewer, not the other way around: if the coach says a player was alone at 0:16, you can see where everybody was at 0:16.
Questions per plan
Free: 5 AI coach queries per month. Pro (US$12/month): 15. Advanced (US$49/month): 60. Each query is a full investigation by the model, with the depth it needs. The diagnosis, the rounds, the viewer and the statistics are available on every plan without using the AI quota, and a local analysis mode answers the standard questions (why lost, what went well, what repeated) with the deterministic engine at no cost.
Frequently asked questions
Which AI model does the coach use?
A large language model from the provider configured by Tacly, used only to write the answer. The facts come from Tacly's own engine, and the answer is fact-checked against the demo before you see it.
Is my demo sent to the AI provider?
Only the tactical data needed to answer the question (player names present in the demo, positions, events) is sent when you ask something. Your e-mail and account data are never sent.
Does it work in Spanish?
Yes. The product, the findings and the coach answer in English or Spanish, following your language setting.
Ask the coach about your last match
Upload a demo, open the coach and ask what you always wanted to know about that round. 5 AI queries a month on the Free plan.
Updated: 2026-11-02