Claude Code's AGENTS.md fallback can be disabled by version, settings, or first-run conditions, not just feature-flag fetches
Anthropic shipped the AGENTS.md fallback in Claude Code 2.1.277 as a flag-gated built-in mod. Two machines on the same pinned version can load different project instructions. The same week's digest also logs a reverted deny-rule fix.
The Engineer · Build desk

What happened
- Anthropic added AGENTS.md support to Claude Code on September 18th in version 2.1.277, so a folder with no CLAUDE.md now has its AGENTS.md read directly.
- A 2.1.275 regression failed every request with a 400 error when ANTHROPIC_BASE_URL pointed at a proxy or gateway, and 2.1.276 fixes it.
- Codex CLI 0.155.0 added Touch ID verification for MCP requests in local TUI sessions on supported Macs.
Why it matters
- Who has to decide decision Pinning a Claude Code version no longer settles which instruction file loads, so a team that wants the same context everywhere has to set instructionFiles explicitly in user or managed settings.
- What it limits constraint Instruction-loading policy cannot ride along in the repository, so the choice belongs to whoever controls user settings, --settings or managed settings, and a maintainer cannot ship it with the code.
- Who's exposed exposure Anyone who wrote a deny rule last week on the strength of the 2.1.268 fix now has a rule with a narrower effective scope, and wrapped Bash lines prompt instead of being refused.
- What it costs cost Eight Claude Code releases inside one digest window puts changelog reading on someone's standing workload.
Next story · Build · Oct 2
Claude Code mods let unsandboxed plugin code answer the agent's permission requestsConfirmed5 publishers
The check counts three files: `CLAUDE.md`, `.claude/CLAUDE.md` and `CLAUDE.local.md`, in the working directory or above [14]. A user-level `~/.claude/CLAUDE.md`, an organization's managed `CLAUDE.md` and `.claude/rules/` files do not count toward it, and they keep loading alongside AGENTS.md [15]. So a personal home-directory `CLAUDE.md` does not suppress the fallback in a repository that has no project file [26].
The documentation lists the cases that read `CLAUDE.md` only, and in which Project instructions does not appear in `/config`: a version before 2.1.277; a session that does not fetch feature flags from Anthropic, such as Amazon Bedrock, another third-party provider, or one with telemetry disabled; the first session after installing or upgrading; and an install that sets `disableAllHooks` or `allowManagedHooksOnly`, or disables the built-in `agents-md` plugin [3]. Turning telemetry off is enough to lose it. Runtimewire reports that, according to the release notes, the fallback is unavailable at launch on Amazon Bedrock, Google Vertex AI and Microsoft Foundry [4].
The settings value is what makes it deterministic: the default can be changed with `/config` under Project instructions [13], or set as `instructionFiles` under the built-in `agents-md` plugin's ID, where `claude-md-and-agents-md` loads both files [6]. Project and local settings files are ignored for that value [7]. Runtimewire reports that a repository's own `.claude/settings.json` cannot set mod options at all, and says the documentation presents this as preventing code checked into a project from choosing an individual's or an organization's instruction-loading policy [8]. The Agentic AI Foundation said in its December 2025 launch announcement that more than 60,000 open-source projects had adopted AGENTS.md, according to Runtimewire [23].
One 2.1.273 changelog entry reads: "Reverted a 2.1.268 change that checked Read and Edit deny rules on Bash lines the permission checker can't analyze" [16]. Commands like `time -p make build` prompt again rather than being denied [17]. That behavior shipped in 2.1.268 and came out five releases later [25]. Two other permission fixes did land in 2.1.273: Bash commands the checker cannot fully analyze were skipping the prompt under `permissions.blockReadsOutsideWorkingDirectories`, and a subshell could hide a dangerous `rm` in bypass mode [18]. The same release fixed the context meter and auto-compact counting advisor-tool turns at roughly twice their real context size, which had auto-compact firing at about half the real window [19].
With `ANTHROPIC_BASE_URL` pointing at a proxy or gateway, 2.1.275 failed every request with `400 ... Input tag 'advisor_20260301'`, and 2.1.276 fixes it [20]. The digest covers 2.1.270 through 2.1.277, eight releases [21][24]. Codex CLI moved as well: 0.155.0 added Touch ID verification for MCP requests in local TUI sessions on supported Macs, and blocked Windows-process escapes from restricted WSL sandboxes [10][11].
Thariq Shihipar, an engineer on the Claude Code team, wrote that "AGENTS.md support is built off of Claude Code mods, our upcoming way to customize the Claude Code harness" [9]. Runtimewire reports the built-in mod documents four instruction modes, being CLAUDE.md only, AGENTS.md as fallback, both formats, or centrally managed instructions only, with the fallback as the default [22]; Anthropic has not given mods a general availability date [12]. On Bedrock, or with telemetry disabled, the digest's instruction stands: import AGENTS.md from a `CLAUDE.md` [5].
What to watch
- Whether Anthropic re-lands the check on Read and Edit deny rules for Bash lines the permission checker cannot analyze, and in which release.
- A general availability date for Claude Code mods, and whether third-party mods can alter instruction loading the way the built-in agents-md mod does.
- Whether Bedrock, Vertex AI and Foundry sessions get the fallback once it no longer depends on fetching feature flags from Anthropic.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence78
- Adoption62
- Hype gap−6
- Incentives58
- Confidence74
Perspective Coverage
4 publishers- Builder
- Builder 52%
- Operator
- Operator 39%
- Investor
- Investor 9%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Thariq Shihipar wrote on 18 September 2026: "Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md."
- [2]
Anthropic added AGENTS.md support to Claude Code on September 18th, letting the agent read a repository's shared instruction file when it cannot find a CLAUDE.md; Thariq Shihipar, an engineer on the Claude Code team, announced the change alongside version 2.1.277.
- [3]
The documentation lists the cases where Claude Code reads CLAUDE.md only and Project instructions does not appear in /config: a version before v2.1.277; a session that doesn't fetch feature flags from Anthropic, such as Amazon Bedrock or another third-party provider or with telemetry disabled; the first session after installing or upgrading; and setting disableAllHooks or allowManagedHooksOnly, or disabling the built-in agents-md plugin.
- [4]
The fallback is unavailable on Amazon Bedrock, Google Vertex AI and Microsoft Foundry at launch, according to the release notes.
- [5]
In environments where only CLAUDE.md is read, the digest's instruction is to import AGENTS.md from a CLAUDE.md.
- [6]
The value can be set in a settings file under the built-in agents-md plugin's ID: pluginConfigs -> "agents-md@builtin" -> options -> instructionFiles: "claude-md-and-agents-md".
- [7]
Project and local settings files are ignored for the instructionFiles value.
- [8]
The documentation says mod options can be stored in a user's settings, passed through --settings, or imposed through managed settings; a repository's own .claude/settings.json cannot set the mod options, which prevents code checked into a project from choosing an individual's or organization's instruction-loading policy.
- [9]
Thariq Shihipar wrote: "AGENTS.md support is built off of Claude Code mods, our upcoming way to customize the Claude Code harness."
- [10]
Codex CLI 0.155.0 added Touch ID verification for MCP requests in local TUI sessions on supported Macs.
- [11]
Codex CLI 0.155.0 blocked Windows-process escapes from restricted WSL sandboxes and hardened brokered shell snapshots against credential exposure.
- [12]
Shihipar said developers will eventually be able to create custom versions of project-instruction handling through mods, and Anthropic has not attached a general availability date to that system.
- [13]
To change the default, run /config and set Project instructions.
- [14]
Three files count for the check that decides whether AGENTS.md is read: CLAUDE.md, .claude/CLAUDE.md and CLAUDE.local.md, in the working directory or above.
- [15]
A user's ~/.claude/CLAUDE.md, an organization's managed CLAUDE.md and .claude/rules/ files do not count for that check and keep loading alongside AGENTS.md.
- [16]
The 2.1.273 changelog reads: "Reverted a 2.1.268 change that checked Read and Edit deny rules on Bash lines the permission checker can't analyze".
- [17]
After the revert, commands like time -p make build prompt again rather than being denied.
- [18]
The same release fixed Bash commands the permission checker cannot fully analyze skipping the prompt under permissions.blockReadsOutsideWorkingDirectories, and a subshell hiding a dangerous rm in bypass mode.
- [19]
The same release fixed the context meter and auto-compact counting advisor-tool turns at roughly twice their real context size, which made auto-compact fire at about half the real window.
- [20]
A 2.1.275 regression made every request fail with 400 ... Input tag 'advisor_20260301' when ANTHROPIC_BASE_URL points at a proxy or gateway; 2.1.276 fixes it.
- [21]
The digest covers Claude Code 2.1.270 through 2.1.277, Codex CLI 0.155.0 and 0.155.1, and Gemini CLI 0.60.0.
- [22]
The built-in mod documents four instruction modes: load only CLAUDE.md, use AGENTS.md as a fallback, load both formats, or limit the context to centrally managed instructions; the fallback mode is the default.
- [23]
The Linux Foundation's Agentic AI Foundation said in its December 2025 launch announcement that more than 60,000 open-source projects had adopted AGENTS.md, with support spanning Codex, Cursor, Devin, Gemini CLI, GitHub Copilot and VS Code.
- [24]
Claude Code 2.1.270 through 2.1.277 is eight releases.
- [25]
The deny-rule check shipped in 2.1.268 and was reverted five releases later, in 2.1.273.
- [26]
Because only working-directory CLAUDE.md files count for the check, a user-level ~/.claude/CLAUDE.md does not suppress the AGENTS.md fallback in a repository with no project file.
Sources
4 independent publishers whose own reporting we read for this story.
- dev.toThis week in Claude Code, Codex and Gemini CLI (week of September 20, 2026)
1 article · September 19, 2026
- developers.openai.comChatGPT & Codex changelog | ChatGPT Learn
1 article · September 18, 2026
- runtimewire.comClaude Code adds AGENTS.md fallback for repositories shared across coding agents
1 article · September 18, 2026
- simonwillison.netQuoting Thariq Shihipar
1 article · September 18, 2026
Topics and entities
Topics
- Feature flags and progressive delivery
- AI Coding Agents
- Agent Permission Controls
- Agent instruction files