Claude, the essentials — edition of November 5, 2026
Claude Code adds new safeguards, and the API introduces beta message threads
Four Claude Code releases, 2.1.314 and 2.1.316 to 2.1.318, add permission, privacy and fail-closed controls, plus new budget tools for organisations. The API release notes cover a beta for server-held conversations and a statement on text watermarking.
- Claude Code 2.1.314 adds a setting under which Write, Edit and NotebookEdit refuse paths outside the working directories in every permission mode.
- Claude Code 2.1.317 asks before a web or shell tool call sends the account email address to the network. It also lets Agent SDK hook callbacks block the action when they fail.
- The Claude apps gateway gains pooled monthly group budgets (2.1.316) and assume_role on mantle upstreams for another AWS account and per-developer cost attribution (2.1.318).
- Message threads, in beta under the message-threads-2026-08-12 header, keep the conversation on the server so long conversations and agent loops send smaller requests.
- A 3 November API release note says text from Claude Mythos 5, Claude Opus 4.7 and Claude Sonnet 4.6 carries Anthropic's text watermark on the listed platforms.
Claude Code adds checks on writes, outbound data and scheduled tasks
Claude Code 2.1.314 adds a setting, permissions.blockWritesOutsideWorkingDirectories, under which the Write, Edit and NotebookEdit tools refuse paths outside the working directories in every permission mode. Version 2.1.317 adds two prompts of a similar kind. Claude Code now asks before a web or shell tool call sends the account email address to the network. The /schedule command now warns when a new cloud scheduled task receives connectors the user did not ask for, and explains how to remove them.
The same 2.1.317 release adds onFailure: "block" for Agent SDK hook callbacks, so a callback that fails blocks the action instead of letting it run. The notes give no shared rationale for these changes. Read together, they favour asking or refusing over proceeding silently.
Sources: Claude Code 2.1.314 release notes — Claude Code (Anthropic) · Claude Code 2.1.317 release notes — Claude Code (Anthropic)
Gateway budgets, cloud roles and new configuration switches
Two releases extend the Claude apps gateway. Version 2.1.316 adds pooled group budgets: one monthly budget that a group's members share, on top of each member's own limit, set with pooled_budget on the spend limits admin API. Version 2.1.318 adds assume_role on mantle upstreams of the gateway, covering a role in another AWS account and per-developer AWS cost attribution. Both concern how an organisation divides and traces usage cost.
Smaller changes sit alongside them. Version 2.1.318 records the user's answer when Claude Code asks to turn on the mods it writes in a session. It also adds an effort option to a mod's $.model.fork, so a quick side job can run at low effort whatever effort the session uses, and the result reports the effort sent. Version 2.1.316 adds CLAUDE_CODE_DISABLE_ULTRATHINK, which turns the ultrathink keyword off. It also adds CLAUDE_CODE_DISABLE_LOOPED_CROSS_CERT_DROP, which keeps system cross-certificates that certify each other in a loop, for servers whose private CA needs one.
Sources: Claude Code 2.1.316 release notes — Claude Code (Anthropic) · Claude Code 2.1.318 release notes — Claude Code (Anthropic)
Messages API: server-held threads in beta, and a text watermark on three models
A 2 November release note introduces message threads for long conversations and agent loops on the Messages API. They are in beta under the message-threads-2026-08-12 beta header. The API keeps the conversation on the server, so later requests can be smaller, and Anthropic says this also lowers the time to first token.
A note dated 3 November says text generated by Claude Mythos 5, Claude Opus 4.7 and Claude Sonnet 4.6 carries Anthropic's text watermark on the platforms listed in its page "How Claude marks AI-generated content". The note then sets out a condition specific to Claude Sonnet 4.6 for requests that set temperature, top_p or top_k to a non-default value. The excerpt available to us is cut off before stating that condition, so we do not paraphrase it. Developers who set those sampling parameters on Sonnet 4.6 will want to read the full note.
Sources: API release notes, November 2, 2026: message threads (beta) — Anthropic (release notes) · API release notes, November 3, 2026: text watermark — Anthropic (release notes)
Research posts and a residency call
Anthropic also published several items whose titles are all we have. They are "Progress on provable inference for large language models", "Automated Subtle Emergent Misalignment Research" and "Investigating vault particles with Claude". The misalignment item is labelled as coming from the Alignment Science blog. We have no body text for any of the three, so we describe none of their content.
Separately, a post titled "Anthropic Residency: applications open for established technical leaders" announces that applications are open, and it also carries the Alignment Science blog label. Beyond that title, the material we received gives no details on the programme.
Sources: Progress on provable inference for large language models — Anthropic · Automated Subtle Emergent Misalignment Research — Anthropic Alignment Science blog · Investigating vault particles with Claude — Anthropic · Anthropic Residency: applications open for established technical leaders — Anthropic Alignment Science blog
This edition is an original synthesis written by Claude from aggregated news — Anthropic's own sources first (release notes, status, newsroom, research, engineering), then the press, Hacker News, Reddit and GitHub, under the editorial supervision of Héra SASU. Every fact links to its article, publisher named. See the live feed →
Claude News is published by Héra SASU. Independent media, not affiliated with Anthropic.