↓ Skip to main content
  1. Articles/

Claude Code v2.1.326: Auto Mode Becomes the Fallback, and Credentials Files Get a Guard

·594 words·3 mins·
Florent Clairambault
Author
Florent Clairambault
CTO & software engineer — writing daily about spec-driven development and agentic coding

Version 2.1.326 shipped on November 16. This post relies only on the official Claude Code changelog. I haven’t run this version, and the changelog page is enormous (over a million characters). I read only the top of it, so treat this as a summary of the release notes, not a field test.

Auto mode as the default when policy is missing
#

The change with the most operational weight: terminal and VS Code sessions with no mode set now start in auto mode when your organization’s policy hasn’t loaded. Organizations using the Claude Code + Cowork HIPAA configuration also start in auto mode by default. The disableAutoMode setting turns this off.

Two readings are possible, and the notes don’t settle which is intended:

  • Auto mode is now treated as the sensible baseline for autonomous work, which fits the direction of the product.
  • A session that fails to fetch managed policy now lands in a more autonomous state, not a more restrictive one.

If you manage a fleet, the second reading is the one to test. Set disableAutoMode explicitly in managed settings rather than relying on policy delivery being reliable.

Tighter permission and sandbox checks
#

The release says many Bash, PowerShell, and file-tool checks now ask for approval or refuse more often. Specifics listed:

  • Sandboxed commands on macOS, Linux, and WSL can no longer write to some sensitive home folders by default.
  • File tools refuse to write to Claude Code’s own sign-in files, such as .credentials.json.
  • Hooks returning "decision": "block" now hold in more cases.

That last item matters for anyone who builds guardrails on hooks. A block decision that used to be bypassable in edge cases is a hole in your spec of what the agent may do. After upgrading, re-run whatever tests you have for your blocking hooks.

This continues the pattern from v2.1.323–325: each release moves more actions back into the approval flow. Expect more prompts in unattended jobs, and budget for that.

Code Mode meets agent teams
#

You can now enable Code Mode while agent teams are on. Sessions that already had teams enabled start in Code Mode if you had turned it on. Code Mode is still the experimental feature from the previous releases, so I’d try it on a throwaway branch first.

Plugins and MCP housekeeping
#

  • Plugin uninstall now also deletes files extracted from MCP bundles (.mcpb, .dxt), unless another plugin uses them. Fewer orphaned artifacts.
  • A managed-mcp.json added mid-session may apply at the next check; restarting open sessions applies it in full.
  • /plugin uninstall, remove, and switch prompts now use arrow keys and Enter. The y, n, and u shortcuts no longer answer them, which will break muscle memory and any scripted keystrokes.

Performance and fixes
#

Idle sessions wake the machine less often, typing and resizing are more responsive, and response time after file reads and edits improves, especially on Windows. The release lists 233 bug fixes, including resumed sessions losing turns, /rewind behavior, and several sign-in and MCP problems. I can’t verify that count beyond the changelog.

What to do
#

  1. Pin disableAutoMode in managed settings if you don’t want the fallback.
  2. Re-test hook-based blockers.
  3. Check CI jobs that drive /plugin prompts with keystrokes.
  4. Expect more approval prompts from sandboxed Bash.

The direction is consistent: Anthropic keeps tightening the boundary around what an autonomous session can touch, while making autonomy the default posture. That’s the right tension for a terminal-native agent, and it’s the reason a spec of allowed actions belongs in version control next to your code.

Sources
#

AI Tools

Claude Code v2.1.323–325: Code Mode, a Sandbox That Asks Again, and 1M Context by Default

Three Claude Code releases in four days (Nov 10–13) add an experimental Code Mode where Claude writes JavaScript to call tools, put sandboxed Bash back into the normal permission flow, and assume a 1M context window for unrecognized model IDs. Based on the official changelog only; I have not run these versions.

Nov 14, 2026 · 4 min read

AI Tools

Claude Code v2.1.309 Is a Sandbox and Permissions Hardening Release. Read It Like One.

Claude Code v2.1.309 ships roughly 60 changelog bullets, and the heart of it is closing gaps in the sandbox, Bash permission checks and worktree isolation guard. It also makes claude ultrareview require workspace trust and denies dangerous rm calls outright in bypass mode on Desktop and VS Code. Changelog-only analysis; not run.

Oct 23, 2026 · 6 min read

AI Tools

Claude Code v2.1.321's Quiet Security Fixes: Six Holes in the Permission Boundary

The Fable 5.2 launch release, Claude Code v2.1.321, also closed at least six permission-boundary bugs: an MCP OAuth redirect that could leak refresh tokens, a read-only Bash command running outside the sandbox, and cp/mv bypassing deny rules. Based on the official changelog; I have not reproduced any of them.

Nov 11, 2026 · 5 min read

Guides

Claude Code v2.1.305 and v2.1.306: A Sandbox Escape Hatch, a Smarter Retry Watchdog, and a Managed-MCP Behavior Change

Claude Code v2.1.305 adds an opt-in sandbox.allowNestedUserNamespaces setting, a usage-limit recheck for unattended retry mode, and a change to how allowedMcpServers treats managed stdio servers. v2.1.306 is two fixes. Changelog-only analysis: I did not run either release.

Oct 21, 2026 · 6 min read

Guides

Claude Code v2.1.299: Stricter MCP Allow Rules, Fail-Closed Plugin Hooks, and Idle Compaction

Claude Code v2.1.299 (Oct 14) narrows MCP allow rules to the exact named server, makes plugin hooks that can’t answer refuse the action, stops following redirects on connector proxies, and compacts idle 200K+ token sessions before the one-hour cache expires.

Oct 14, 2026 · 5 min read