Anthropic drops DMCA notice on developer reverse-engineering Claude Code

Anthropic just gave developers a textbook lesson in how not to handle developer relations.

It started when a programmer decided to poke around under the hood of Claude Code, Anthropic’s terminal-based coding assistant. Because Anthropic obfuscated the tool's source code, the developer reverse-engineered it, cleaned it up, and posted the source on GitHub.

Anthropic moved fast, slapping the repo with a formal DMCA takedown notice to get it scrubbed. Anthropic did not respond to requests for comment.

A tale of two developer toolkits

The move sparked immediate blowback across social media, largely because OpenAI is playing this game with the exact opposite playbook.

Both Claude Code and OpenAI’s Codex CLI serve roughly the same purpose: they let programmers lean on frontier models from the command line to write, test, and debug software. Both dropped within months of each other in a fierce sprint for developer mindshare.

Codex CLI launched under a permissive Apache 2.0 license, allowing anyone to modify, inspect, or commercially redistribute the code. Claude Code, on the other hand, is shackled to a restrictive commercial license that explicitly forbids modifications without permission.

To make matters worse for Anthropic’s optics, OpenAI has spent the last week actively merging dozens of community pull requests into Codex CLI. One of those merged patches actually lets Codex CLI query rival models — including Claude itself.

To be fair to Anthropic, Claude Code is still in an early, somewhat buggy beta, and companies often obfuscate code for basic security hygiene. The lab might eventually choose to open source it down the line.

Why it matters

Developer loyalty is won in the terminal, and terminal users despise closed black boxes.

For engineers, this contrast makes Codex CLI much more attractive right now. Coders don't just want a smart autocomplete; they want hackability, local control, and the freedom to wire in alternative model providers when rate limits hit or bills climb.

We think Anthropic stumbled hard on this one. Reaching for copyright takedowns against a developer experimenting with client-side code looks defensive and tone-deaf, especially in an open-source-dominated environment.

It is also an unexpected PR coup for OpenAI. After years of moving toward closed-door proprietary releases — prompting CEO Sam Altman to admit earlier this year that OpenAI had been on the "wrong side of history" regarding open source — Codex CLI proves they know how to charm the engineering crowd when they want to.

When frontier models are neck and neck on coding benchmarks, ecosystem goodwill and licensing freedoms can easily tip the scales.