Chess Analytics

Queue an engine analysis (Stockfish plus move comments and a report) of one game: a chess.com or Lichess game of the linked account, given by the `url` or `id` from `get_dashboard`, or pasted PGN text. Returns { id, status } immediately; an analysis that already exists comes back `done`. Poll `get_analysis` with the id until its status is `done`. Premium plan; limited to 30 games per hour.

analyze_game

Queue engine analyses for the user's most recent games that have none yet (1 to 20, newest first), optionally filtered by source and time class. Returns per game the id and status; poll `get_analysis` with an id for progress. Games that were already analyzed are skipped. Premium plan; each game counts toward the limit of 30 per hour.

analyze_recent_games

Write an AI coach review of the user's latest 10 games for this source and time class (patterns across games, strengths, what to work on). All 10 games need a finished analysis first: otherwise nothing is written and status `missing` lists the games to pass to `analyze_game`. Returns the stored review if the same games were already reviewed. Uses a language model. Premium plan; limited to 5 per hour.

create_coach_review

One engine analysis of one game. While queued or running it returns status, progress and the queue position and estimate; call again later. When done it returns the players, opening, accuracy and move-quality counts per side, accuracy per phase, the written report and the key moments (worst mistakes and best moves) with the engine's preferred move and any move comment. With `detail` = `full` the text also lists every move. The structured result carries the complete analysis (every position and move comment) for the interactive game review view. Analyses are created with `analyze_game`.

get_analysis

The most recent stored AI coach review of the user's latest 10 games for this source and time class: the report, which analyses it covers and whether newer games exist since. Status `none` means no review has been created yet (`create_coach_review` creates one). Reads what is stored; it doesn't call the language model.

get_coach_review

Results and trends over the user's most recent chess.com and/or Lichess games: win/draw/loss, score by site, time class and colour, rating changes, openings, how games ended, and a newest-first game list. Each game row has its `url` and, when analyzed, an `analysis_id` for `get_analysis`. Free plans get these statistics from the sites' own data; with Premium, games that have an engine analysis add accuracy and engine-derived stats (`engine_stats`). Fetches the game lists from chess.com and Lichess.

get_dashboard

The signed-in user's account: username, plan (free or premium) and the linked chess.com and Lichess usernames. Plan and linked accounts determine what the other tools return: Premium is required for engine analyses and AI reviews; without a linked account the game tools have no games to show.

get_profile

One multi-game trend report: the period compared with the previous one, engine-derived stats (accuracy, phases, openings, error patterns), the written report with focus areas and the games it covers. Returns progress while the report is still being built.

get_aggregate

The user's engine-analyzed games, newest first: id, status (queued, running, done, failed), players, result, date, perspective and accuracy. Use an id with `get_analysis`. Lists what exists; it doesn't start analyses.

list_analyses

The user's multi-game trend reports (weekly, monthly, ...): id, period label, time class, source, status, number of games and accuracy. Use an id with `get_aggregate`. Lists what exists; it doesn't create reports.

list_aggregates