io.github.darshjoshi/pitwall

io.github.darshjoshi/pitwall

Turn Claude into your F1 race engineer.

Category
Visit Server

README

<!-- mcp-name: io.github.darshjoshi/pitwall --> <h1 align="center">Pitwall</h1>

<p align="center"> <strong>Turn Claude into your F1 race engineer.</strong><br> Real telemetry. Real strategy data. Real-time during races. 75 years of history. </p>

<p align="center"> <a href="https://pypi.org/project/f1pitwall/"><img src="https://img.shields.io/pypi/v/f1pitwall?color=orange&label=PyPI" alt="PyPI"></a> <a href="https://pypi.org/project/f1pitwall/"><img src="https://img.shields.io/pypi/dm/f1pitwall?color=blue&label=Downloads" alt="Downloads"></a> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-green.svg" alt="License: MIT"></a> <a href="https://www.python.org/downloads/"><img src="https://img.shields.io/badge/Python-3.10+-blue.svg" alt="Python 3.10+"></a> <a href="https://modelcontextprotocol.io"><img src="https://img.shields.io/badge/MCP-Compatible-purple.svg" alt="MCP Compatible"></a> <a href="https://github.com/darshjoshi/pitwall/stargazers"><img src="https://img.shields.io/github/stars/darshjoshi/pitwall?style=social" alt="GitHub Stars"></a> </p>

<p align="center"> <img src="https://raw.githubusercontent.com/darshjoshi/pitwall/main/assets/ver_vs_nor_abu_dhabi_2024_quali.png" alt="Verstappen vs Norris — Abu Dhabi 2024 Qualifying Speed Trace" width="700"> </p>


Quick Start

pip install f1pitwall
claude mcp add pitwall -- f1pitwall

Then ask Claude: "Who won the 2025 Australian GP?"

Want visual plots and deep analysis? pip install "f1pitwall[full]"

<details> <summary><strong>Install from source instead</strong></summary>

git clone https://github.com/darshjoshi/pitwall.git && cd pitwall
pip install "mcp[cli]" requests        # lite
pip install -r requirements-full.txt   # full
claude mcp add pitwall -- python3 $(pwd)/pitwall.py

</details>


Why Pitwall?

Claude knows F1 from training data — but it can't look up last week's race. It can't show you Verstappen's throttle trace through Turn 1. It doesn't know who pitted first or when the safety car came out.

Pitwall connects Claude to live F1 data:

  • Real data, not hallucinations — actual timing feeds from formula1.com
  • Lap-level telemetry — speed, RPM, throttle, brake, gear, DRS at 4Hz per car
  • Visual plots — speed trace comparisons, gear shift maps returned as images
  • 75 years of history — every race result and championship since 1950
  • Live during races — real-time positions, gaps, weather, and race control
  • Zero API keys — all core data is free, no account needed

What You Can Ask

"Who won the 2025 Australian GP?"           → Race results and classification
"Verstappen's speed on lap 25 at Monaco"    → Lap telemetry at 4Hz
"Plot Hamilton vs Norris speed trace"       → Visual speed comparison chart
"Compare Ferrari's tyre strategy"           → Stint-by-stint breakdown
"Who won the 1994 championship?"            → 75 years of history
"When was the safety car at Silverstone?"   → Race control messages and flags

<details> <summary><strong>See all example questions</strong></summary>

Question Tool Used
"Who won the Chinese GP?" get_standings
"What was Verstappen's speed on lap 25?" get_telemetry
"Compare Hamilton vs Leclerc" get_driver_comparison
"What tyres did everyone use?" get_tyre_strategy
"Fastest pit stop at Australia 2025?" get_pit_stops
"When was the safety car?" get_race_control
"Was it raining during the race?" get_weather
"Top speeds at Monza 2024?" get_speed_traps
"Norris's lap times in the race" get_lap_times
"Who won the 2005 championship?" get_championship_standings
"Plot Verstappen vs Hamilton speed trace" plot_telemetry_comparison
"Show me the gear shift map at Monaco" plot_gear_shifts
"Who gained the most positions?" compare_grid_to_finish
"Overtakes in the race" detect_overtakes
"Compare Verstappen lap 5 vs lap 50" plot_multi_telemetry_comparison
"Ferrari head-to-head in qualifying" team_head_to_head
"Deleted laps in qualifying" get_deleted_laps
"Gap to leader throughout the race" get_gap_to_leader

</details>


Features

67 tools across two modes. Pitwall auto-detects what's installed — no config changes needed.

Lite Mode (14 tools)

pip install "mcp[cli]" requests — no heavy dependencies.

Race results, lap times, telemetry, tyre strategy, pit stops, weather, race control, speed traps, driver comparison, and historical data back to 1950. Uses F1's free static archive and the Jolpica API.

Full Mode (67 tools)

pip install -r requirements-full.txt — adds FastF1.

Everything in Lite, plus:

Category What You Get
Visual Plots Speed trace comparisons, gear shift maps, multi-lap telemetry overlays
Deep Telemetry Brake point analysis, RPM patterns, DRS usage, throttle traces
Advanced Strategy Stint degradation, compound comparisons, tire age performance
Race Intelligence Overtake detection, gap tracking, position changes, qualifying progression
Live Data Real-time positions, lap times, sector times, weather during active sessions

<details> <summary><strong>Full tool list (67 tools)</strong></summary>

Lite Tools (always available)

Tool Description
list_seasons Available seasons (2018-present)
list_races Full season calendar with dates
get_race_info Session details and available data feeds
get_standings Race classification — positions, gaps, best laps, pits
get_lap_times Lap-by-lap times, filterable by driver and lap range
get_telemetry Speed, RPM, throttle, brake, gear, DRS for a specific lap
get_tyre_strategy Compound, stint length, new/used for every driver
get_pit_stops All pit stops sorted by fastest
get_race_control Flags, penalties, safety cars, investigations
get_weather Air/track temp, rain, humidity, wind
get_speed_traps Speed at 4 measurement points per driver
get_driver_comparison Head-to-head: position, pace, strategy, pit stops
get_historical_results Race results from 1950 to present
get_championship_standings Driver/constructor championships from 1950+

FastF1 Tools (requires FastF1)

Category Tools
Visual Plots plot_telemetry_comparison, plot_gear_shifts, plot_multi_telemetry_comparison, plot_driver_telemetry_comparison
Telemetry Analysis analyze_brake_points, analyze_rpm_data, analyze_drs_usage
Lap Analysis get_lap_times_fastf1, get_deleted_laps, analyze_lap_consistency, get_fastest_sectors, get_personal_best_laps, compare_sector_times
Strategy get_driver_tyre_detail, get_stint_analysis, compare_tire_compounds, compare_tire_age_performance, analyze_starting_tires, compare_strategies
Race Analysis get_race_results, get_sprint_results, get_session_summary, get_fastest_lap_data, detect_overtakes, compare_grid_to_finish, get_qualifying_progression
Pit Stops get_pit_stop_detail, get_fastest_pit_stops
Driver & Team get_driver_info, get_driver_standings, get_constructor_standings, team_head_to_head, get_team_laps, analyze_long_run_pace
Track & Safety get_circuit_info, get_track_status, get_track_record, get_race_control_messages, get_penalties, get_dnf_list
Speed & Position get_speed_trap_comparison, get_position_changes, get_gap_to_leader
History get_race_winners_history
Live Data get_live_session_status, get_live_positions, get_live_lap_times, get_live_sector_times, get_live_telemetry, get_live_weather
Session get_schedule, get_session_info, get_weather_data

</details>


Setup

<details> <summary><strong>macOS</strong></summary>

Claude Code:

claude mcp add pitwall -- python3 /absolute/path/to/pitwall.py

Claude Desktop — add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "pitwall": {
      "command": "python3",
      "args": ["/absolute/path/to/pitwall.py"]
    }
  }
}

</details>

<details> <summary><strong>Windows</strong></summary>

1. Find your Python path:

where python

This will return something like C:\Users\YourName\AppData\Local\Programs\Python\Python313\python.exe or C:\Python313\python.exe.

2. Note where you cloned Pitwall: For example: C:\Users\YourName\Projects\pitwall\pitwall.py

Claude Code (PowerShell):

claude mcp add pitwall -- python C:\Users\YourName\Projects\pitwall\pitwall.py

Claude Desktop — add to %APPDATA%\Claude\claude_desktop_config.json:

To open this folder, press Win + R, type %APPDATA%\Claude, and hit Enter. If the Claude folder or claude_desktop_config.json doesn't exist, create them.

{
  "mcpServers": {
    "pitwall": {
      "command": "python",
      "args": ["C:\\Users\\YourName\\Projects\\pitwall\\pitwall.py"]
    }
  }
}

Note: Use double backslashes (\\) in the JSON path, or forward slashes (/) — both work. The command is python (not python3) on Windows. </details>

<details> <summary><strong>Linux</strong></summary>

Claude Code:

claude mcp add pitwall -- python3 /absolute/path/to/pitwall.py

Claude Desktop — add to ~/.config/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "pitwall": {
      "command": "python3",
      "args": ["/absolute/path/to/pitwall.py"]
    }
  }
}

</details>

Restart Claude Code or Claude Desktop after setup. Works with any MCP-compatible client.

Optional: Beginner-Friendly Skill (Claude Desktop)

Upload SKILL.md as a skill in Claude Desktop (Settings → Skills → Upload). Claude will explain F1 jargon inline — DRS, undercut, compound, safety car, etc.


Data Sources

All core data is free and requires no API keys.

Source Coverage What it provides
F1 Static Live Timing 2018-present Telemetry, timing, strategy, pit stops, weather, race control
Jolpica-F1 1950-present Historical results and championships
FastF1 (optional) 2018-present Enhanced telemetry analysis and visual plots
F1 SignalR Core (optional) Live only Real-time race data during active sessions

How It Works

Pitwall reads from F1's publicly available static timing archive — the same data that powers the official F1 app. After each session ends (~30 minutes), F1 publishes 33 data feeds per session including full car telemetry (speed, RPM, throttle, brake, gear, DRS at ~4Hz per car), GPS positions, tyre data, pit stops, and race control messages.

The telemetry tool (get_telemetry) correlates the timing stream with the car data stream to extract telemetry for a specific driver on a specific lap — something no other F1 MCP server does.

Architecture

Claude ──MCP──> Pitwall ──HTTP──> livetiming.formula1.com/static/ (free)
                       ──HTTP──> api.jolpi.ca/ergast/f1/ (free)
                       ──lib──>  FastF1 (optional, local)
                       ──WS───> SignalR Core (optional, live races)

Running the Server

python3 pitwall.py              # MCP stdio (Claude Code / Claude Desktop)
python3 pitwall.py --http       # MCP HTTP (remote / self-hosted)
python3 pitwall.py --http --port 3000

Live Race Data

Pitwall includes a raw SignalR Core WebSocket client for real-time data during active F1 sessions. Most data is free — car telemetry and GPS require an F1 TV Pro or Premium subscription.

<details> <summary><strong>What's free vs what needs F1 TV</strong></summary>

Data Free F1 TV
Race positions, gaps, lap times Yes Yes
Race control, flags, penalties Yes Yes
Weather, track status Yes Yes
Tyre compounds, stint info Yes Yes
Team radio URLs Yes Yes
Car telemetry (speed, RPM, throttle, brake) No Yes
GPS positions (X/Y/Z coordinates) No Yes

All data (including telemetry and GPS) becomes free in the static archive ~30 minutes after a session ends.

</details>

<details> <summary><strong>Authentication setup</strong></summary>

python3 auth_setup.py

This opens a browser for F1 TV login. The token is saved locally:

  • <project_dir>/.f1token
  • ~/Library/Application Support/fastf1/f1auth.json (macOS)

Token expires every ~4 days. Re-run to refresh. Never uploaded anywhere.

</details>

<details> <summary><strong>Live client usage</strong></summary>

import asyncio
from signalr_client import F1LiveClient

async def main():
    # Free mode — timing, weather, race control (no auth needed)
    client = F1LiveClient(no_auth=True)

    @client.on("TimingData")
    def on_timing(data, timestamp):
        for num, info in data.get("Lines", {}).items():
            print(f"P{info.get('Position','?')} #{num} Gap: {info.get('GapToLeader','')}")

    @client.on("RaceControlMessages")
    def on_rc(data, timestamp):
        for msg in data.get("Messages", {}).values():
            print(f"[{msg.get('Flag', '')}] {msg.get('Message', '')}")

    await client.connect()

asyncio.run(main())

For full telemetry (speed, RPM, throttle, brake, GPS):

from auth_setup import load_token

client = F1LiveClient(no_auth=False, auth_token=load_token())

@client.on("CarData.z")
def on_telemetry(data, timestamp):
    # Speed, RPM, throttle, brake, gear, DRS at ~4Hz per car
    ...

@client.on("Position.z")
def on_position(data, timestamp):
    # GPS X/Y/Z coordinates at ~4Hz per car
    ...

</details>


Reference

Race Names

Race names are fuzzy-matched. All of these work:

"china", "chinese", "shanghai"           → Chinese Grand Prix
"australia", "melbourne", "aus"          → Australian Grand Prix
"monaco", "monte carlo"                  → Monaco Grand Prix
"silverstone", "great britain", "british" → British Grand Prix

Driver Codes

VER = Verstappen    HAM = Hamilton    NOR = Norris     LEC = Leclerc
ANT = Antonelli     RUS = Russell     PIA = Piastri    BEA = Bearman
GAS = Gasly         LAW = Lawson      HAD = Hadjar     SAI = Sainz
ALO = Alonso        STR = Stroll      OCO = Ocon       BOT = Bottas
ALB = Albon         HUL = Hulkenberg  COL = Colapinto  LIN = Lindblad

Project Files

File Purpose
pitwall.py MCP server — 67 tools, auto-degrades to 14 without FastF1
signalr_client.py Raw SignalR Core WebSocket client for live race data
decompressor.py Zlib decompression for CarData.z / Position.z
merger.py Keyframe + delta state management for F1's incremental format
topics.py All 20 SignalR topics with auth/compression metadata
auth_setup.py F1 TV token setup — browser-based OAuth flow

Contributing

Found a bug? Want to add a tool? Contributions are welcome.

  1. Fork the repo
  2. Create a feature branch
  3. Make your changes
  4. Run the test suite: python3 tests/pitwall_tool_validation.py
  5. Open a pull request

Credits

  • FastF1 by @theOehrly — the gold standard F1 Python library
  • Jolpica-F1 — the Ergast API successor
  • drivenrajat/f1 — inspiration for FastF1 tool patterns

Built by

Darsh Joshi — AI Engineer

LinkedIn GitHub Email

License

MIT

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured