BattleLM is a native macOS app for running multiple AI coding agents side by side, chatting with them one-on-one, and letting them discuss problems together in group chats.
It currently focuses on:
- Native macOS desktop experience
- Local AI CLIs such as Claude, Gemini, Codex, and Qwen
- Group-chat workflows for comparing answers and debating solutions
- iPhone pairing for remote access to your AI instances
- Create multiple AI instances with separate working directories
- Chat with a single AI in a clean 1:1 view
- Create group chats and let multiple AIs respond to the same problem
- Choose model variants, including second-level reasoning effort where supported
- Pair your Mac with iPhone through a QR flow
- macOS
- Xcode for local development
- At least one supported AI CLI installed and logged in
Supported providers in the current app:
- Claude
- Gemini
- Codex
- Qwen
BattleLM works with locally installed AI CLIs. Typical examples:
- Claude:
claude - Gemini:
gemini - Codex:
codex - Qwen:
qwen
If a CLI is installed but not authenticated yet, BattleLM will detect that and show it as not fully ready.
- Open
BattleLM.xcodeprojin Xcode. - Select the
BattleLMscheme for macOS. - Build or Archive from Xcode.
The repository also includes an iPhone companion target:
BattleLM-iOS
- Add one or more AI instances.
- Choose a working directory for each instance.
- Start with a 1:1 chat or create a group chat.
- Compare responses, iterate, and switch models when needed.
- Pair with iPhone if you want mobile access to the same remote host.
Releases are intended to ship notarized macOS builds.
Current release assets use the macOS DMG format, for example:
BattleLM-macOS-1.0.3.dmg
BattleLM is currently centered on the native macOS app and its iPhone companion. The old cross-platform desktop packaging line is no longer part of the project.