How a Remote Feature Flag Silently Broke Claude Code's AGENTS.md for Privacy-Conscious and Enterprise Users
Here's what happened. On September 18, Claude Code added support for AGENTS.md, the cross-tool instruction file already read by 30+ coding agents. Big announcement. Real gap closed.
Except it didn't actually work for a meaningful chunk of users.
An independent engineer ran a simple test: drop a unique canary word into an AGENTS.md file, ask Claude Code to recite it back. Then vary the environment.
What they found:
• Telemetry disabled? File silently skipped.
• Running on Amazon Bedrock, Google Vertex AI, or Microsoft Foundry? Silently skipped.
• Behind a corporate gateway that blocks "nonessential" traffic? Silently skipped.
• No warning, no error, no log line — the session just ran without your instructions.
The root cause: reading a local markdown file had been wired to a remote feature flag. If the client couldn't reach Anthropic's servers to check whether the feature was "on," it defaulted to off. A capability that should need zero network access ended up depending on one.
The people most affected weren't random — they were exactly the users most likely to have adopted AGENTS.md in the first place: teams standardizing one instruction file across multiple AI coding tools, privacy-conscious developers who turn telemetry off on purpose, and enterprises on managed cloud platforms.
The workaround was almost insulting in its simplicity: a one-line CLAUDE.md file containing @AGENTS.md, which uses Claude Code's existing import mechanism and never depended on the flag at all.
Five days after launch — the same day the investigation was published — Anthropic shipped v2.1.281, removing the gate for telemetry-disabled sessions, Bedrock, Vertex, Foundry, and LLM gateways.
What I find most instructive here isn't the bug — it's the pattern. The same two environment variables were already known to gate an unrelated feature (the post-session feedback survey), and that gating was documented and intentional. This one wasn't. The difference between "designed behavior" and "silent side effect" often comes down to whether anyone thought to test the edge case before shipping.
Worth remembering next time you're staging a feature behind a flag: closed-by-default fallbacks are fine for gradual rollouts to the general population. They're much riskier for capabilities that a fixed, non-shrinking subset of your users structurally cannot reach.
If you're running Claude Code, this is a good moment to check claude --version and confirm you're on v2.1.281 or later.
Full piece on mer.vin — link posted separately in comments.
This is accurate, well-documented, and matches multiple independent sources closely — including a GitHub issue, release notes, and technical write-ups covering the same incident.
The timeline checks out precisely. AGENTS.md support shipped in Claude Code v2.1.277. A bug report was filed against it on September 20, 2026, and the fix landed in v2.1.281 on September 23, 2026 — consistent with your "five days after launch" framing if the feature launched around September 18.
The core bug is confirmed exactly as described, including the mechanism. Independent coverage confirms AGENTS.md loading was implemented as a plugin gated behind a remote feature flag (tengu_agents_md_mod, served by Statsig) rather than built into the normal file loader. To check whether the flag was on, Claude Code had to reach that service — and if it couldn't, the plugin registered as unavailable and the file was never read. That's your "wired to a remote feature flag, defaulted to off" description, confirmed precisely.
Every affected scenario in your bullet list is independently confirmed, including via a documented before/after table: telemetry disabled (CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1), running through Bedrock or Vertex, and no error surfaced when it happened.
The workaround is confirmed word for word. Multiple sources give the identical fix: put one line in your CLAUDE.md — @AGENTS.md — since CLAUDE.md's import mechanism doesn't depend on the feature flag at all.
The fix in v2.1.281 is confirmed directly from release documentation, stating: AGENTS.md now works on Amazon Bedrock and with telemetry disabled — a prior limitation that blocked enterprise/air-gapped users is gone, with the feature-flag-fetch requirement removed entirely for Bedrock and third-party-provider sessions.
One detail I couldn't independently verify: the specific claim that the same two environment variables were already known to gate the post-session feedback survey as documented, intentional behavior. This is a plausible and specific insider detail consistent with everything else checking out, but I didn't find independent confirmation of it.
One thing worth adding, since it strengthens rather than undercuts your point: this wasn't an isolated incident. A related, still-open GitHub issue shows the same telemetry-disable flag also silently broke an unrelated feature (Channels/plugin messaging) with no clear error — meaning the pattern of "opting out of telemetry silently disables unrelated gated features" is a recurring architectural issue at Claude Code, not a one-off AGENTS.md bug. That's a stronger version of your closing point about closed-by-default fallbacks being risky for structurally-unreachable user segments.
Nothing here is exaggerated — if anything, the pattern is more systemic than the post suggests.