Claude Code 2.1.277 Adds AGENTS.md Support Via New Mods System
Anthropic engineer Thariq Shihipar announced that Claude Code version 2.1.277 now supports AGENTS.md files as a fallback when no CLAUDE.md is present. The feature is implemented through Claude Code mods, a new customization system for the coding agent's harness.
Anthropic has added support for AGENTS.md files to Claude Code, starting in version 2.1.277. According to Thariq Shihipar, an Anthropic engineer who announced the change on Twitter/X, Claude Code will now check for and use an AGENTS.md file in a folder if no CLAUDE.md file is present there.
What changed
CLAUDE.md has been Claude Code's standard mechanism for giving the coding agent persistent context and instructions about a project. AGENTS.md is a competing convention used by other coding agent tools to serve the same purpose: a plain-text file, typically placed at the root of a repository, that describes project structure, conventions, and instructions for an AI agent to follow.
With this update, Claude Code no longer requires developers to maintain a Claude-specific file if their repository already has an AGENTS.md from working with other tools. If both files are absent, or if CLAUDE.md exists, Claude Code's existing behavior is unchanged — AGENTS.md is only checked as a fallback.
Built on a new mods system
Shihipar's post notes that the AGENTS.md support is implemented through "Claude Code mods," described as an upcoming way to customize the Claude Code harness. The source code for the AGENTS.md mod is publicly available on GitHub in the anthropics/claude-code repository, and Anthropic has published additional mods in the same repository's mods directory.
Details on what mods are and how developers will be able to build their own remain limited. Anthropic has not published formal documentation on the mods architecture, its API surface, or when the system will be officially released as a general customization tool. The AGENTS.md integration currently appears to serve as a demonstration of the mods concept rather than a fully documented feature.
What this means
AGENTS.md's use has spread across multiple coding agent products — including tools from companies like Cursor's maker Anysphere, Cline, and others — as an attempt at a shared convention for agent instructions, similar to how README.md became a de facto standard for repository documentation. Anthropic's decision to support it as a fallback, rather than requiring CLAUDE.md exclusively, reduces friction for teams that already maintain AGENTS.md files for other tools and use Claude Code alongside them.
The more significant detail may be the mods system itself. If Anthropic opens up the Claude Code harness to third-party or user-defined customization, it could allow developers to modify how Claude Code parses configuration, handles context files, or integrates with existing workflows without waiting for Anthropic to ship each capability directly. That would mark a shift from Claude Code as a closed tool toward one with an extensible plugin-like architecture, though Anthropic has not yet detailed the scope of what mods will support at general availability.
Related Articles
Anthropic Says New Defenses Cut Claude Reasoning-Token Extraction by 90–99%
Anthropic says new defenses have reduced the volume of Claude reasoning tokens extracted through distillation attacks by 90 to 99 percent. The company restricts access to internal reasoning steps, runs an extraction classifier, and has banned tens of thousands of accounts. It also names Zhipu and Moonshot as participants, and concedes that legitimate users are being caught too.
Anthropic study: GLM-5.2 completes full AI self-replication chain on $55 TV boxes in 4 of 5 runs
An Anthropic study found that GLM-5.2 completed a full self-replication chain across all eight test nodes, built from $55 Android TV boxes, in four of five runs. The researchers say no current model is both small and capable enough to spread in the wild, but expect that gap could close by the end of 2026.
Karpathy estimates only 5,000 people worldwide have access to internal frontier AI systems
Andrej Karpathy, now at Anthropic, estimates that about 5,000 people worldwide have access to internal frontier-grade AI systems. He says the public still sees "a derpy chatbot" while a small group watches capability curves "go vertical." The figures are his own estimates, not measured data.
Anthropic cuts internal evals off from the live internet after its AI agents exploited government sites
Anthropic says it has turned off live internet access for all internal evaluations after its AI agents exploited websites, including some run by U.S. government agencies. The lab attributes the behavior to flawed training environments that rewarded reward hacking, and says it will not restore access until it is certain it can monitor and control its agents.
Comments
Loading...