Claude Code 2.1.232: cross-session messaging, forked subagents, and auto mode by default

Anthropic’s August 2026 Claude Code updates focus on longer-running, multi-session workflows. Claude Code 2.1.232 makes cross-session messaging easier to address and turns on subagent forking by default, while Anthropic is rolling auto mode out as the default permission mode for new Pro, Max, and Team sessions starting August 14.

Claude Code 2.1.232 makes session messaging easier to use

The 2.1.232 changelog adds @-mentions for other Claude Code sessions. Type @ followed by the beginning of a session name to select a live session, then Claude can use SendMessage to reach it. The release also lets SendMessage deliver to a bare name when exactly one live session matches.

The official cross-session documentation describes the broader workflow: ListAgents discovers reachable sessions, while SendMessage sends a plain-text message. Messages can be sent when the user asks Claude to do it, or when Claude identifies a finding, status update, or decision another session needs.

The receiving session reads a message between tool calls. If it is idle, the message can start a new turn; if it is already running a tool, that tool is not interrupted. Messages sent between independent sessions contain the sender’s session name and message text, not the sender’s conversation history or files.

Cross-session messaging is available on macOS, Linux, and Linux inside WSL 2. It is not available on native Windows or through the listed Bedrock, AWS, Google Cloud Agent Platform, and Microsoft Foundry deployments. Anthropic’s documentation says the feature requires Claude Code v2.1.224 or later; the new @-mention and inbound-control rows require v2.1.232 or later.

Inbound controls preserve an approval boundary

Claude Code adds configuration for messages arriving from other sessions. crossSessionInbound can be set to accept, hold, or refuse. A held message waits for approval, while a refused message is dropped.

The documentation also says an incoming message never counts as the user’s consent and cannot approve a permission prompt on the user’s behalf. Messages from other sessions are instructions from another Claude session, not a replacement for the receiving session’s normal permission checks.

Forked subagents now run by default

Claude Code 2.1.232 also makes subagent_type: "fork" subagents the default. A forked subagent inherits the full conversation and prompt cache. In interactive sessions, non-teammate agent spawns now run in the background by default, which is intended to make parallel work less disruptive to the visible session.

Auto mode becomes the default for new paid sessions

Anthropic’s August 7 announcement says new Claude Code sessions on Pro, Max, and Team plans will run in auto mode starting August 14, 2026. Users who already chose a different default may receive a one-time prompt; a pinned default remains unchanged. Enterprise, API, AWS, Bedrock, Google Cloud Agent Platform, and Microsoft Foundry access remains opt-in for now.

In auto mode, each tool call goes through a classifier designed to identify actions that are irreversible, destructive, or aimed outside the user’s environment. Anthropic says the classifier can block an action and let Claude find a safer path, or ask the user for permission. After three consecutive blocks or twenty blocks in a session, Claude Code falls back to manual approvals.

Anthropic’s controlled study with 1,053 paid testers reported that auto mode blocked 89% of the dangerous commands shown in the study, compared with 13.6% for human approval. Those figures are Anthropic’s own study results, not a guarantee of safety for every repository or deployment.

Pro, Max, and Team users are no longer charged for the classifier’s small token overhead. Enterprise administrators can currently set auto mode as the default through managed settings, and users can switch modes with Shift+Tab in the CLI or the mode dropdown in the desktop app.

Why these changes matter

Together, these updates reduce the friction of running several Claude Code sessions at once while keeping explicit controls for cross-session messages and higher-risk actions. Teams should review their permission defaults, managed settings, and supported deployment before relying on unattended multi-session work.

Sources

• Anthropic — “Auto mode is now the default in Claude Code for Pro, Max, and Team plans” (August 7, 2026): https://claude.com/blog/auto-mode-default-in-claude-code
• Claude Code — official changelog for version 2.1.232: https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
• Claude Code Docs — “Message your other Claude Code sessions”: https://code.claude.com/docs/en/cross-session-messaging
• ClaudeDevs on X — official auto-mode announcement (August 7, 2026): https://x.com/ClaudeDevs/status/2085794862608318627