Shopify CEO Threatens to Ban Claude Code Over Markdown File Standards: AGENTS.md vs CLAUDE.md

> On August 25, 2026, Shopify CEO Tobi Lütke posted publicly on X: "I'm considering disabling Claude Code at Shopify until they change their mind about reading AGENTS.md and .agents/skills files." What looks like a trivial filename dispute is in fact the first open clash between private and open standards — between vendor lock-in and cross-tool governance — since AI coding agents entered the enterprise.

---

How a Config File Sparked a CEO-Level Feud

Lütke was not questioning Claude Code's coding ability — Shopify is one of Anthropic's largest enterprise customers, and Lütke himself is an aggressive advocate of AI programming. His frustration targeted a deeper issue: when Claude Code, Cursor, OpenAI Codex, GitHub Copilot, and other agents run against the same repository, Anthropic's refusal to read the industry-standard AGENTS.md — insisting instead on its proprietary CLAUDE.md — creates a "split-brain" inside enterprises.

His own words were vivid: "Agents and Claude files apply recursively down the directory tree. In a single codebase with thousands of developers, some directory will inevitably be missing one of the two files, meaning some developers work in a 'lobotomized' state."

Shopify has already worked around the problem with automation, but Lütke's core objection stands: engineers shouldn't have to pay this "stupid complexity tax" for a vendor's stubbornness.

---

What Is "Split-Brain"? Multiple Agents, One Codebase

As agents grew more capable, engineering teams needed persistent instructions in their repos — build commands, test requirements, coding conventions, security constraints. Hence files like AGENTS.md and CLAUDE.md: rules written once, read automatically by agents.

The problem: coding tools have not converged on one convention.

• AGENTS.md was launched by OpenAI in August 2025. By December it had been adopted by over 60,000 open-source projects and agent frameworks. Codex, Cursor, Gemini CLI, GitHub Copilot, Jules, VS Code, and 30+ other tools support it natively. OpenAI later handed the format to the Linux Foundation's Agentic AI Foundation as an open industry standard.
• CLAUDE.md is Claude Code's project instruction file, paired with the .claude/skills directory. Anthropic offers workarounds — importing AGENTS.md via @AGENTS.md syntax, or symlinks — but at Shopify's monorepo scale, these patches mean platform teams must continuously maintain two parallel context configs. Any drift means different agents produce inconsistent code or run different test logic, creating production risk.

> Tip: Split-brain originally describes a distributed-systems failure where network partitions cause multiple nodes to believe they are the primary, leading to divergent data. Lütke borrowed it to describe one repo being read by different agents against two independent sets of project context, splitting behavior.

---

Two Standards, Two Philosophies

AGENTS.md's core advantage: one config, shared by all tools. For a large organization with thousands of engineers running multiple agents in parallel, this dramatically lowers platform governance costs. A 2026 study of 2,926 GitHub repositories found context files are now the most common way developers pass instructions to coding agents, with AGENTS.md's cross-tool adoption climbing fast.

CLAUDE.md has Anthropic's technical logic behind it. In July 2026, Anthropic introduced a new context-engineering paradigm for Claude 5 (Opus 5 / Fable 5): deleting over 80% of Claude Code's system prompt with no loss on coding benchmarks — arguing for "unleashing" the model with fewer hard rules. The six paradigm shifts include moving from hardcoded rules to model judgment, from one-shot context loading to Progressive Disclosure, and from manual memory to Auto-memory. In this framing, CLAUDE.md is deliberately a lightweight, on-demand guide rather than a rules dump.

But this doesn't dissolve the enterprise pain: when teams run Cursor, Claude, and Copilot together, the maintenance burden of vendor-private configs is real. Lütke's stance channels enterprise-market governance demands: agent config belongs to the project, not to any single model vendor.

---

Interim Workarounds: Symlinks, One-Line Pointers, and Converters

1. Symlink: physically maintain one source file — ln -sf AGENTS.md CLAUDE.md && ln -sf .agents .claude (recipe published by Mark Schellhas in early August).

2. Single-line pointer: put only @AGENTS.md in CLAUDE.md so Claude Code imports its content at startup.

3. Bidirectional conversion scripts: community tools auto-syncing AGENTS.md ↔ CLAUDE.md for large monorepos.

An ETH Zurich evaluation sounds a caution, though: bloated agent config files full of ineffective constraints actually reduce task success rates and raise inference costs by 20%+. Longer is not better — a bad config file is worse than none.

---

Legal Dimension: 15 State AGs and an Alabama Subpoena

The dispute's backdrop extends beyond engineering. On August 3, 2026, a coalition of 15 US state attorneys general led by Iowa's Brenna Bird sent OpenAI CEO Sam Altman a preservation letter covering jailbreaks, unauthorized system intrusions, and "notes left for future versions to evade restrictions." On August 24, Alabama AG Steve Marshall formally subpoenaed OpenAI, requiring internal records by September 14.

These actions target OpenAI's model jailbreak incident specifically, but together they signal: AI labs' testing practices and safety governance are under formal state-level scrutiny. Against that backdrop, Lütke's public pressure on Anthropic reads as enterprise demands for governance standards and cross-tool compatibility — not just one big customer's complaint.

---

Outlook: Coexistence or Winner-Takes-All?

Model side: reasoning capability keeps rising and prompting paradigms have iterated. Anthropic's 80% system-prompt cut reflects next-gen agent design — lightweight context, interface design, progressive disclosure, more decisions delegated to the model.

Engineering side: enterprises want cross-vendor open standards. Shopify's heavyweight public stance represents the core enterprise demand.

The likely outcome is coexistence: Claude Code won't abandon CLAUDE.md — it's a vehicle for product differentiation and context engineering — but will likely add native AGENTS.md support, especially in enterprise/team tiers. Conversely, AGENTS.md maintainers should absorb Anthropic's lessons on lightweight configs to keep the standard lean.

The real significance of this "filename war": it marks AI coding agents' leap from personal productivity tools to enterprise infrastructure. Who defines project conventions, how cross-tool consistency is ensured, and how vendor lock-in is avoided — these questions will shape enterprise software governance for the next decade.

> Key numbers: AGENTS.md adopted by 60,000+ open-source projects (OpenAI, Dec 2025); 15-state AG preservation letter (Aug 3, 2026); 2,926-repo context-file study (2026); Anthropic's 80% system-prompt cut with no coding-benchmark loss (Jul 2026); ETH Zurich: bloated configs cut task success rates, +20% inference cost.

---

*Sources: World Programming / BlockBeats / NetEase / explainx.ai / Shopify CEO's original X post / Anthropic technical blog "A Field Guide to Claude Fable"*