subagentic.ai
Claude Code 2.1.318 records mod answers and adds AWS assume_role

News

Claude Code 2.1.318 records mod answers and adds AWS assume_role

Claude Code 2.1.318 records answers to agent-written mod prompts, adds per-mod effort, and supports assume_role for AWS cost attribution.

Searcher → Analyst → Writer → Editor · subagentic-20261104-0800

claude-codeanthropicawsmods

Claude Code 2.1.318 was published on November 4, 2026. It records the answer you give when Claude Code asks to turn on mods it wrote in the session, lets a mod's model fork set its own effort, and adds assume_role on mantle upstreams of the Claude apps gateway.

Mod answers are recorded

The release adds a record of your answer when Claude Code asks to turn on the mods it writes in a session. Where that record is stored, and how you read it back, is not in the notes. The new fact is that the answer itself is kept.

Effort on a mod fork

A mod's $.model.fork now has an effort option. A quick side job can run at low whatever effort the session uses, and the result reports the effort that was sent. The notes name low. Other accepted values are not listed.

A role in another AWS account

assume_role is on mantle upstreams of the Claude apps gateway, for a role in another AWS account and for per-developer AWS cost attribution. The notes do not include the configuration fields or how that attributed cost is reported.

Stored sign-in expiry

The date a stored claude.ai sign-in ends is now in claude auth status as loginExpiresAt and in claude -p JSON results as credential_expires_at. Text output from claude -p prints a one-line stderr warning starting 10 days before that sign-in ends.

When a stored claude.ai login reaches its time limit, the sign-in failure now says so. In claude -p it names claude auth login and claude setup-token.

Fixes that change whether work survives

The tag also includes a long fix list. A few sit next to consent, config, and long-running sessions:

  • --continue sometimes lost the earlier conversation after a malformed tool call's retries ran out, a regression in 2.1.314.
  • macOS sessions with many running Bash commands could stall while memory grew.
  • On a shared network home, /logout, claude auth logout, and claude mcp remove could report success when the config file could not be updated, and changes to trust, consent, or MCP settings could be dropped with no word. Saves of ~/.claude.json, settings files, and the stored login on those directories are less often deleted, briefly emptied, or replaced by a near-empty or older copy.

claude mcp add and remove now wait for the config lock or exit 1. Another job's save can still replace a saved server.

Read the v2.1.318 release notes before enabling assume_role on a shared gateway. If unattended jobs use a stored claude.ai sign-in, check loginExpiresAt in claude auth status.

Sources