Brifing

Anthropic Adopts OpenAI's AI Agent Instruction Standard

Claude now supports rival's AGENTS.md format to simplify developer workflows

1Format unification move

Anthropic's Claude Code will now support OpenAI's AGENTS.md format when CLAUDE.md isn't present, eliminating the need for duplicate instruction files. This change applies to version 2.1.277 and can be configured via the /config command. Developers previously maintained separate Markdown files for each platform, specifying agent behavior, tool usage, and coding conventions. The formats were similar but contained platform-specific variations, forcing complex workarounds like symlinks to keep them synchronized.

2Developer workflow impact

The change significantly reduces maintenance overhead for developers using both Claude and OpenAI tools. Previously, teams had to manage multiple versions of instruction files - both platform-specific and project-specific variations. These documents form the foundational context for AI coding agents, read with every request. The unification means developers can now standardize on AGENTS.md while retaining the option to use Claude-specific instructions when needed through the toggle feature.

3OpenAI's standardization push

OpenAI contributed AGENTS.md to the Agentic AI Foundation under the Linux Foundation last year to promote adoption. By December 2025, over 60,000 open source projects had implemented the standard. Anthropic's decision to support it marks a significant endorsement, given Claude Code's popularity in the developer community. This suggests the market is consolidating around common standards despite initial platform-specific approaches.

4Strategic cooperation signal

The move represents a rare moment of cooperation between competing AI platforms. OpenAI's head of core product publicly welcomed the decision on social media, exchanging positive emojis with an Anthropic engineer. While the technical impact is modest, the symbolic significance is notable in an industry often characterized by proprietary approaches. It suggests pragmatism may be prevailing over specification separatism in developer tools.

5Implementation details

Claude will now check for AGENTS.md when CLAUDE.md isn't present in a folder, using the same file structure developers already employ for OpenAI tools. The behavior is configurable, allowing teams to opt out if needed. This backward-compatible approach ensures existing workflows aren't disrupted while providing a path to standardization. The change reflects Anthropic's confidence in its platform's adoption, no longer feeling compelled to maintain completely separate standards.

6Community reaction

Prominent software developers widely praised the decision on social media, suggesting the pain point of maintaining multiple instruction formats was significant. The positive reception indicates that while the technical change is small, its impact on developer experience is substantial. Many viewed it as an example of the kind of interoperability the industry needs more of, especially in AI tooling ecosystems.

7Market implications

Anthropic's move suggests the competitive dynamics in AI developer tools may be shifting from differentiation through proprietary formats to competing on core capabilities. With over 60,000 projects using AGENTS.md, supporting the standard became more valuable than maintaining separate specifications. This could signal a broader trend toward standardization in AI tooling interfaces, similar to what occurred in other software development ecosystems.

8Future standardization potential

This development may pave the way for further interoperability between AI coding tools. While the change currently affects only instruction file formats, it establishes a precedent for cooperation on standards. As the AI tooling market matures, we may see more consolidation around common interfaces, reducing lock-in and making it easier for developers to work across multiple platforms without excessive overhead.

Original article: www.theregister.com - Articles — Anthropic Adopts OpenAI's AI Agent Instruction Standard