MCP server exposing race telemetry read-only tools for Formula 1 session analysis, including lap data, telemetry, strategy, incidents, and replay functionality.
26 tools with consistent verb-noun naming and comprehensive descriptions (avg 150 chars). All tools have input schemas with typed parameters and descriptions. Output schemas are documented in descriptions but not formally specified in JSON Schema. Error handling is implicit (read-only operations). Tool composition is strong, tools chain well (e.g., list_sessions → get_session_drivers → get_driver_laps). Key gaps: no explicit output schema definitions, no enum constraints for string parameters (sessionType, eventType, groupBy accept free-form strings), and no pagination guidance for large result sets.
Aggregate telemetry statistics: group by driver/lap/stint/compound/track-status, compute metrics like avg speed, brake time, DRS active time.
Per-driver stint analysis: lap times, tyre degradation, pace trends, and compound performance.
Pit stop analysis: stop duration, lap loss, tyre change, and strategic impact.
Distance-domain lap comparison: aligned telemetry by track distance with delta and segment analysis.
Narrative comparison of two laps: delta, sector breakdown, and segment-by-segment advantage.
Detect telemetry event windows: hard braking, high speed, DRS active, throttle lift. Returns time ranges and driver context.
String parameters accept free-form values instead of enums. sessionType, eventType, groupBy, metrics, and channels lack enum constraints, inviting hallucinated invalid values.
Output schemas are described in natural language but not formally specified in JSON Schema. LLMs cannot reliably parse unstructured text responses or plan downstream tool calls.
No pagination parameters (limit, offset, cursor) or result count guidance. Tools returning large arrays (get_positions, get_driver_laps, aggregate_telemetry) risk context window exhaustion.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2025-06-18+ | v2 |
Comprehensive race debrief: all sections (overview, strategy, incidents, weather) in one response. Everything-at-once mega-tool.
List non-deleted laps for a driver in a session.
Identify and analyze braking zones in a lap: entry speed, peak brake, duration, and comparison to other drivers.
Data quality metrics for one lap: telemetry coverage, interpolation percentage, gaps, and validation status.
Narrative summary of one lap: pace, tyre, strategy context, and telemetry insights.
Get bounded telemetry samples for one driver's lap.
Overtakes and position changes: driver, lap, position delta, and context.
Lap-by-lap position changes: driver, position, gap to leader, and lap number.
Race control messages timeline: flags, incidents, safety cars, and regulatory actions.
Narrative race summary: overview, key moments, strategy, incidents, and weather. Slimmed version; use focused tools for details.
Get a time-bounded replay chunk with position and telemetry for all drivers.
Get narrative context for a replay window: weather, track status, race control messages, and pit activity.
Get replay bounds, available drivers, available channels, track markers, context availability, and weather summary.
List drivers for one imported session.
Headline pre-counted facts for a session: driver count, total laps, safety-car/red-flag/VSC deployments, fastest lap, top speed, peak track temp, and whether it rained. Prefer this for 'how many / who / what was the' questions instead of counting other tools' arrays. Returns a plain-text sentence.
Final race standings: position, driver, team, points, gaps, and laps completed.
Weather timeline: temperature, humidity, rainfall, and wind across the session.
List imported sessions. Defaults to race sessions; pass another session type only when explicitly needed.
Search telemetry events by type, driver, lap range, and distance range. Returns matching samples with context.
Strategy summary: pit stops, tyre compounds, stint lengths, and undercut/overcut opportunities.
No error recovery guidance. Tools return implicit failures (empty results, null fields) without actionable error messages or suggestions for next steps.
Tool composition relies on ToolBundleRouter keyword matching rather than explicit tool selection hints. LLMs may select wrong bundles for ambiguous queries.