FirstLight_CR
FirstLight_CR is a research toolkit that trains and tests Clash Royale AI players offline, using replay learning and self-play, for developers and AI researchers.
What it is good for
- Training an AI to play Clash Royale from replays
- Running offline Clash Royale matches against a trained model
- Comparing two game-playing AI models head to head
At a glance
- Who it is for
- Setup difficulty
- Developer ยท Requires programming
Level 5 of 5. You need to write code to use it.
- Cost
- Free and open source
- Type
- Framework or SDK
- Works with
- MuMu Player, Hugging Face, Null's Royale, PyTorch
How to get started with FirstLight_CR
- Install Windows, Python 3.12 with Tkinter, and an Android 12 MuMu Player instance with root and ARM64 support.
- Create a virtual environment and install
requirements.txt, then copy.env.exampleto.envand fill in the settings. - Install and open the original game APK in MuMu, let it update, then keep it as
user-provided/nulls-royale.apk. - Run
./check_original_game.ps1and./build_offline_engine.ps1to build the offline APK. - Open the interface with
./launch_interface.cmdfrom the repository root.
From the project's own documentation. If a step does not match what you see, the official pages have the latest instructions.
Before you try it
- You must supply the supported game APK yourself; the repository does not include it.
- Desktop setup needs Windows, Python 3.12, and an Android emulator with root.
- Building the APK from source needs JDK 17, Android NDK, and SDK build-tools.
- The project is independent and not affiliated with Supercell.
Momentum
- Rank this week
- #30 in builder tools
- Best rank
- #30 week 46, 2026
- Stars gained last week
- +58
- GitHub stars
- 306
GitHub numbers as of Nov 16, 2026. Updated weekly.
Ranking history
Official links
Questions
- Is FirstLight_CR free?
- Yes. It is free and open source (Apache-2.0 license).
- Do I need to know how to code to use FirstLight_CR?
- Yes. You need to write code to use it.
- What does FirstLight_CR work with?
- According to its own documentation, it works with MuMu Player, Hugging Face, Null's Royale, and PyTorch.
- How do I get started with FirstLight_CR?
- 1. Install Windows, Python 3.12 with Tkinter, and an Android 12 MuMu Player instance with root and ARM64 support. 2. Create a virtual environment and install requirements.txt, then copy .env.example to .env and fill in the settings. 3. Install and open the original game APK in MuMu, let it update, then keep it as user-provided/nulls-royale.apk. 4. Run ./check_original_game.ps1 and ./build_offline_engine.ps1 to build the offline APK. 5. Open the interface with ./launch_interface.cmd from the repository root.
- What should I know before trying FirstLight_CR?
- You must supply the supported game APK yourself; the repository does not include it. Desktop setup needs Windows, Python 3.12, and an Android emulator with root. Building the APK from source needs JDK 17, Android NDK, and SDK build-tools. The project is independent and not affiliated with Supercell.
- What is FirstLight_CR good for?
- Training an AI to play Clash Royale from replays, running offline Clash Royale matches against a trained model, and comparing two game-playing AI models head to head.
Similar agents to FirstLight_CR
- AgentDescent
Improves prompts and agent setups from your own examples rather than training game-playing agents.
- feynman
Research agent for literature reviews from the command line, not a game training toolkit.
- Khoj
Personal AI app for answering questions from documents, with no reinforcement learning or game simulation.
- Firecrawl
Firecrawl is a web data API that searches, scrapes, and interacts with websites to give AI agents clean page content for developers.
- TradingAgents
TradingAgents is a framework that runs a team of AI analysts, researchers, and traders to evaluate stocks and propose trades for researchers and developers.
- Scrapling
Scrapling is a Python web scraping framework that collects data from simple or JavaScript-heavy sites, with an MCP server for AI agents, for developers.