Fastdata

Discover fights and players in a Warcraft Logs report

Use this first when the user provides a Warcraft Logs URL or report code. The MCP server fetches report metadata through its configured WCL proxy. Do not scrape the WCL web page and do not search the local workspace for fixtures. When interaction.state=needs_user_selection, show the available fights and ask the user to choose fightId and playerId. Never choose for the user unless they explicitly request automatic or representative selection.

Analyze a Warcraft Logs fight with WoWAnalyzerCN

Fetch the user-selected fight/player through the configured WCL proxy and run the pinned WoWAnalyzerCN parser. Only call after the user explicitly selected fightId and playerId, unless they explicitly requested automatic selection. The server waits briefly and returns ready when possible, otherwise a pollable analysisId. When interaction.state=awaiting_user_question, ask what the user wants to inspect; do not automatically fan out to findings, metrics, timeline, or capabilities. IMPORTANT: When

Check WCL analysis status

Long-poll a wcl_analyze_fight job for a bounded interval. When interaction.state=awaiting_user_question, ask the user what they want to inspect. Do not automatically call snapshot query tools.

List findings from snapshot

Page through upstream findings stored in the snapshot. Call only when the user asks for findings or an overall triage. A zero-length result means WoWAnalyzerCN emitted no findings; it does not prove perfect play. Does not trigger the Parser. IMPORTANT: When showing a finding to the user, use `description` for the concrete data summary and `evidencePreview[].summary` for each evidence event (e.g. "1.23s: 玩家10 施放 寒冰箭 → 玩家200"). Do NOT show the raw `evidenceRefs` event IDs (ev-1, ev-39) — those are

Get single finding with metrics and evidence

Fetch one finding by ID, with its referenced metrics and evidence event IDs. Does not trigger the Parser. IMPORTANT: When showing the finding, use `finding.description` for the concrete data summary and `finding.evidencePreview[].summary` for each evidence event. Do NOT show the raw `evidenceRefs` / `evidenceEventIds` event IDs (ev-1, ev-39) — those are internal handles. `evidencePreview` is already inlined so you do NOT need to call wcl_get_evidence just to render evidence text; only call wcl_g

Get module metrics from snapshot

Page through upstream metrics stored in the snapshot. Call only when the user asks about a metric or module. Filter by moduleKey. Does not trigger the Parser.

Get timeline entries from snapshot

Query timeline entries by moduleKey and/or timestamp range. Call only when the user asks about timing or sequence. Does not trigger the Parser.

Get event-graph evidence slice

Fetch a slice of the event graph by evidenceRefs or pagination. Returns events + links among them. Does not trigger the Parser.

Quick triage of top findings

Select the top N upstream findings by performance (fail first). Call only after the user explicitly asks for an overall triage or findings. Optional focus filter. Does NOT run custom diagnostics — only selects from upstream findings (P0 §7.3). IMPORTANT: When showing a finding, use `description` for the concrete data summary and `evidencePreview[].summary` for each evidence event. Do NOT show raw `evidenceRefs` event IDs (ev-1, ev-39) — those are internal handles. `evidencePreview` is already in

Deep dive on a topic from snapshot

Organize snapshot content for a topic explicitly requested by the user. Does NOT run custom analysis or thresholds (P0 §7.4). IMPORTANT: When showing a finding, use `description` for the concrete data summary and `evidencePreview[].summary` for each evidence event. Do NOT show raw `evidenceRefs` event IDs (ev-1, ev-39) — those are internal handles. `evidencePreview` is already inlined so you do NOT need to call wcl_get_evidence just to render evidence text.

Show engine capabilities

Return the engine manifest, input manifest, and capability manifest for an analysis. Call only when the user asks about parser coverage, provenance, or whether zero findings reflect missing data. Does not trigger the Parser.