A new open-source tool called Coder is quietly reshaping how developers orchestrate AI coding agents across different providers. The CLI, shared on Hacker News July 11, creates a unified bridge between Anthropic's Claude Code and OpenAI's Codex (plus ChatGPT hosts), allowing these powerful but siloed systems to hand off tasks to each other without context bleed or subscription lock-in.

How Cross-Engine Delegation Actually Works

The tool operates through a common CLI interface paired with standardized skills that both platforms can understand. When Claude Code encounters a subtaskโ€”whether it's writing tests, refactoring a module, or spinning up boilerplateโ€”it can spin up a Codex agent to handle it independently. The inverse works too: Codex agents can push work back to Claude when the task better suits its strengths. This bidirectional delegation keeps each engine's context window uncluttered because you're not dumping an entire project into one prompt.

Why Subscription Spreading Matters

For power users already paying for both Claude and Codex subscriptions, this isn't just about convenienceโ€”it's about economics and capability optimization. Different models excel at different things. Some developers prefer Claude's reasoning style for complex architectural decisions while leaning on Codex for rapid code generation. Coder lets you route work to the right engine without manually copy-pasting context between ChatGPT, Claude.ai, and their respective APIs.

The Skill-Based Architecture

The implementation relies on skill definitions that both platforms can invokeโ€”a clever abstraction layer that sidesteps proprietary agent protocols. This means the tool isn't tied to specific version numbers or API quirks; as both providers evolve, the skills layer can adapt. It's a pattern other tooling developers should watch: rather than building against unstable internal APIs, create portable skill contracts.

Key Takeaways

  • Coder enables bidirectional task delegation between Claude Code and Codex/ChatGPT hosts
  • The CLI plus skills architecture keeps context windows clean across both platforms
  • Developers can optimize for model strengths rather than forcing everything through one subscription
  • Skill-based abstraction provides a future-proof bridge as AI coding tools evolve