The Lab Notes

Anthropic

Claude Code warns Pro and Max users before a resume costs usage

Claude Code 2.1.304 warns Pro and Max subscribers when resuming an inactive conversation would use a noticeable share of the 5-hour usage limit.

The Lab Notes ·

A white hourglass with orange sand tilted over a purple, pink and orange stopwatch, set against a split background: a purple gradient with thin white arcs on the left and a black panel patterned with small hourglass and stopwatch icons on the right.

Claude Code now warns Pro and Max subscribers before they resume an inactive conversation that would use a noticeable share of the 5-hour usage limit. The prompt lets them resume anyway or start a new conversation instead.

The warning is in Claude Code 2.1.304, according to the changelog at code.claude.com. The npm registry lists the release as published Oct. 19, 2026 at 19:47 UTC, after 2.1.303 on Oct. 18.

A New Conversation Can Still Reach the Old One

Claude can look up the previous conversation from a new one if the user refers to it, according to the changelog.

Also in the Release

A new /auto-mode-setup command walks users through teaching auto mode about their environment. It reads the project and recent sessions. It now works on Bedrock, Vertex, Foundry, the Claude apps gateway and sessions with telemetry disabled.

The /claude-api managed-agents-secure command compares a Managed Agents agent’s settings with Anthropic’s published security guide. It changes nothing. Names of tools called earlier now appear in the compaction summary in the transcript view, opened with Ctrl+O.

claude --worktree could remove the worktree and its branch at exit without asking when HEAD was detached or on another branch, which could lose commits. The release fixes that.

Running rm -r on a mounted drive or a top-level folder such as /Volumes, /media or WSL’s /mnt/c did not trigger the safety prompt that rm -rf /Applications does. That is fixed.

Sandboxed commands also failed to start on Linux when a sandbox.filesystem.denyRead glob contained brackets that are not a character class, such as a [user-id] route folder. The release fixes that as well.

Our View

We think the useful detail is that the warning offers a new conversation as one of its two choices. The 5-hour limit is the cap on how much a Pro or Max plan can use in a 5-hour stretch. A conversation that uses a noticeable share of it leaves less for everything else in that stretch.

Pro and Max subscribers see that cost before they resume, but only on resumes that would use a noticeable share.

The benefit depends on subscribers choosing the new conversation when they do not need the old one’s details.