AI scouting for esports · running in production
Weekly opponent scouting reports, generated from raw match data
Esports teams receive a ready-to-train preparation package every week. The Claude API reads demos, championship stats and head-to-head history, and every claim in the report is backed by a number from the raw data.
What the team gets every week
Opponent tendencies
Map pool, pacing, economy patterns and composition preferences — pulled from the opponent's full match history, not from memory.
Cost of each pattern
The report states what each tendency costs against your roster, with the number that backs it: round win rates, first-contact outcomes, utility usage.
Training plan
The preparation ends with what to drill this week, ready for the practice session. Coaches review it in minutes, not evenings.
Claude runs the core pipeline via the API: tool use against our match database, long-context reads of full opponent histories, structured output for the report tables.
How it works
1 · Ingest
Demos and official championship stats flow into our match database as the single source of truth.
2 · Analyze
Claude queries the database through tool use, reads the opponent's complete history in context, and drafts the report with per-claim citations.
3 · Deliver
Every number is checked against the raw data before the report reaches the coach. The audit trail stays attached to each claim.
Pricing
Starter
- 1 team, weekly report
- CS2 or Valorant
- Full audit trail
TeamMost picked
- 1 team, weekly + per-match reports
- Both games included
- Coach review session monthly
Org
- Up to 5 rosters
- Custom integrations
- Priority pipeline slots
Thylor Labs was founded in August 2026 and is bootstrapped. Four partner teams run their weekly preparation on our reports today, with 60+ reports delivered — each one checked against the raw match data before it ships.