라이벌이 한 팀으로? AI 코딩의 새 시대 OpenAI가 Anthropic의 Claude Code 안에서 실행되는 Codex 플러그인을 출시했습니다. 경쟁을 넘어 개발자 편의를
agents anthropic claude openai
| Source: Mastodon | Original article
OpenAI has unveiled a Codex plug‑in that runs inside Anthropic’s Claude Code, effectively letting the two rival AI‑coding agents operate as a single development assistant. The plug‑in, announced on OpenAI’s blog on 31 March, embeds the Codex model—OpenAI’s long‑standing code‑generation engine—within Claude Code’s agentic workflow, allowing developers to invoke either model from the same terminal‑style interface.
We first covered Claude Code in depth on 1 April with “Claude Code Unpacked: A visual guide” (see our earlier report). Since then the tool has become the flagship of Anthropic’s AIAgent era, offering file‑level edits, command execution and context‑aware suggestions. By integrating Codex, OpenAI is not merely licensing a model; it is granting Claude Code access to Codex’s extensive training on public repositories and its fine‑tuned ability to generate concise snippets for a wide range of languages. The result is a hybrid assistant that can switch between Claude 3.5 Sonnet’s conversational reasoning and Codex’s raw code synthesis on the fly.
The partnership matters for three reasons. First, it blurs the line between competing AI ecosystems, signalling a shift from siloed offerings to collaborative tooling that prioritises developer convenience. Second, it could reshape pricing dynamics: OpenAI’s pay‑per‑use Codex may now be bundled into Anthropic’s consumption‑based plans, potentially lowering the barrier for small teams. Third, the combined agent sets a new benchmark for AI‑augmented IDEs, challenging Microsoft’s Copilot and other emerging plugins to match the breadth of integrated capabilities.
What to watch next: OpenAI and Anthropic have promised a public beta in early May, with performance metrics against standalone Claude Code and Codex slated for release. Developers will be keen to see latency, token‑cost comparisons and how the plug‑in handles conflict resolution when the two models suggest divergent solutions. A broader rollout to cloud IDEs such as GitHub Codespaces and JetBrains Fleet could cement the collaboration as a de‑facto standard for AI‑driven coding. Subsequent announcements—especially around pricing tiers or additional third‑party integrations—will reveal whether this joint venture marks the beginning of a more open AI‑coding marketplace or a one‑off strategic experiment.
Sources
Back to AIPULSEN