Phoenix 24
The confrontation concerns interoperability, not the coding assistant’s technical performance.
OTTAWA, CANADA
Shopify chief executive Tobi Lütke says he is considering prohibiting Claude Code inside the company unless Anthropic adds native support for AGENTS.md and related agent-skill files. No ban has yet been implemented. His warning targets the way the coding assistant receives project instructions across Shopify’s extensive software repository rather than the quality or security of Claude’s programming capabilities.
Files such as AGENTS.md provide persistent directions to AI coding tools, including testing requirements, architectural rules and formatting conventions. Their instructions can change across folders within a repository, allowing an agent to receive increasingly specific guidance as it moves through a project. Claude Code instead relies principally on Anthropic’s CLAUDE.md format and does not automatically interpret every competing convention.
For a small development team, maintaining both files or linking one to the other may be manageable. Shopify operates a large monorepository used by thousands of engineers and several AI assistants. Lütke argues that when one tool fails to read the same instructions as the others, developers may unknowingly generate code under inconsistent rules. He described the resulting loss of context as a form of “lobotomy” for the affected agent.
Anthropic has suggested workarounds, including importing AGENTS.md content into CLAUDE.md or creating symbolic links between the files. Shopify has also developed automation to synchronize instructions. Lütke maintains that these measures impose an unnecessary “complexity tax” on enterprise users. Feature requests for broader native compatibility have reportedly been closed as not planned, although Anthropic has not announced a policy change in direct response to his warning.
The dispute exposes a larger contest over standards in AI-assisted software development. As autonomous coding tools become embedded in corporate workflows, proprietary instruction formats can lock organizations into particular vendors or force them to maintain parallel configurations. Shopify’s leverage could encourage greater interoperability, but the episode also demonstrates that adopting AI at scale depends on governance and shared context as much as model intelligence. A powerful agent that reads the wrong rules can create risk faster than an ordinary developer.
Detrás de cada dato, la intención. / Behind every data point, the intention.