Daily brief

Gemini CLI Can List Its Models Without Opening a Chat, Claude Code Asks Before Sending Your Email, and Copilot Moves Into Xcode

By · 5 Nov 2026

A developer seen from behind at a birch desk looks at two monitors, one with a grid of nine lavender tiles and one with a single wide coral tile, in front of a red wall.

At a glance

  • Gemini CLI can now list its models without starting a chat session.
  • Gemini CLI also stops leftover tool work after you cancel a non-interactive run.
  • Claude Code now asks before a tool sends your account email.
  • Copilot is in public preview inside Xcode's own coding assistant.

The briefs for 2 to 4 November did not go out, so today covers the official items from that gap rather than backfilling those dates as their own posts. Gemini CLI v0.67.0 reached the npm latest tag on Wednesday afternoon, Claude Code moved latest twice between Monday and Tuesday evening, and GitHub posted Copilot for Xcode on Wednesday and JetBrains Air on Tuesday. A Gemini CLI preview build landed the same afternoon as v0.67.0, but the stable release leads.

Treat today as a check your defaults day: list the models Gemini CLI actually offers, move Claude Code to v2.1.318 and read claude auth status, and try Xcode's built in assistant on a test Mac before replacing the extension.

Top Stories

1. Gemini CLI v0.67.0 lists its models

Google published Gemini CLI v0.67.0 at about 3:27 PM ET on Wednesday, 4 November, and the npm latest tag moved to it at 3:32 PM ET. The headline for scripts and CI is model discovery. The merged pull request in this release adds gemini models, gemini models list and a top level gemini --list-models flag. With --output-format json or -o json, the output is a JSON array with modelId, id, name, displayName, description and contextLimits, which holds inputTokens and outputTokens.

The same build bridges the IDE companion into container sandboxes over exec stdio. Cancellation is tighter too. In non interactive mode the CLI now handles termination signals and prevents tool execution after you cancel, and it stops continuation stream processing and the automatic nudge after a request is cancelled. Relative file paths now resolve against the active IDE workspace root, and IDE settings and GEMINI.md files stay out of open file context.

A v0.68.0-preview.0 build published about 2:49 PM ET the same afternoon on the npm preview channel. Keep it on a test machine only. The default today is v0.67.0.

Practical dev impact: If a script hard codes a Gemini model name, run gemini --list-models after you move to v0.67.0 and check that the name is still there, and expect a cancelled run to stop calling tools instead of finishing its last step.

2. Claude Code asks before sharing your email

Claude Code v2.1.317 reached npm at 6:57 PM ET on Monday, 2 November, and v2.1.318 followed at 4:55 PM ET on Tuesday, 3 November. Both are stable, and v2.1.318 is the current npm latest. The official changelog dates the entries 3 and 4 November.

The privacy change in v2.1.317 is the one most teams will feel. Claude Code now asks before a web or shell tool call sends your account email address to the network. Our read: an unattended run may now pause here. The same release adds a warning to /schedule when a new cloud scheduled task picks up connectors you did not ask for, with how to remove them, and Agent SDK hook callbacks gain onFailure: "block", so a callback that fails blocks the action instead of letting it run.

v2.1.318 is the release scheduled jobs have needed. claude auth status now shows loginExpiresAt, the date a stored claude.ai sign in ends, and claude -p JSON results carry the same date as credential_expires_at. Starting 10 days before that date, claude -p text output also prints a one line warning on stderr. The build fixes --continue sometimes losing the earlier conversation after the retries for a malformed tool call ran out, a regression from v2.1.314. Claude Code also no longer installs or updates the IDE extension when started in an IDE terminal. Use /ide from an outside terminal or your editor's extensions view instead.

Practical dev impact: Move to v2.1.318 and run claude auth status once, because the email prompt can pause an unattended run, and the expiry date tells you whether a stored login will outlast your next scheduled job.

3. Copilot joins Xcode's coding assistant

On 4 November, GitHub said you can now work with Copilot directly in Xcode's built in coding assistant. It is a public preview, available starting with Xcode 27.3 beta, and you need a Mac running macOS Tahoe 26.6 or later. You set it up in Xcode under Settings, then Intelligence. All Copilot plans are supported, and the existing GitHub Copilot for Xcode extension remains a separate option, so nobody is forced to switch.

A day earlier, GitHub put Copilot in JetBrains Air. That integration uses the Agent Client Protocol and runs tasks in your JetBrains IDE or in the cloud on your existing Copilot subscription. Cloud tasks go through Air Teams and run in a Docker container on a fresh virtual machine, and they use your Copilot plan rather than JetBrains AI credits.

Neither item changes Copilot CLI. The npm latest tag for @github/copilot is still v1.0.105, and a v1.0.106 prerelease sits on its own channel for test machines only.

Practical dev impact: If someone on your team already runs Xcode 27.3 beta on Tahoe 26.6 or later, let them try the built in assistant on that Mac and keep the extension everywhere else, and if you live in JetBrains Air, pick Copilot in the agent picker so the work draws on your Copilot plan instead of JetBrains AI credits.

Practical Impact Analysis

The through line today is knowing what a tool will do before it does it. Gemini CLI can now tell a script which models exist and how large their context limits are, so wrappers stop guessing at model strings. Claude Code asks before your email leaves the machine and warns well ahead of a stored login running out, which matters most for anyone running claude -p from cron or CI. Our read: the email prompt is the change most likely to surprise an unattended job, so test one scheduled run by hand before trusting it overnight.

Elsewhere, Cursor did not move. Its main changelog still leads with the 20 October Agent SDK entry, and its CLI changelog still leads with v2026.10.22. Codex latest is still v0.166.0, and Qwen Code v0.26.7 arrived Tuesday night with an operator scoped model.catalogRefresh setting. If you only have time for one pass, move Gemini CLI to v0.67.0 and Claude Code to v2.1.318, list Gemini's models and read Claude's auth status.

Tutorial

  1. Open a new terminal on a test machine. Nothing below installs anything or changes files.
  2. Step 1 prints the npm tags for Gemini CLI and Claude Code and the versions you have installed.
  3. Step 2 holds the optional pinned installs, all commented out. The preview build is for a test machine only.
  4. Step 3 lists Gemini models and saves nothing. If the flag is missing, you are not on v0.67.0 yet.
  5. Step 4 prints Claude Code's auth status as JSON. If there is no loginExpiresAt field at all, you are not on v2.1.318 yet. If it reads null, the field is there but no claude.ai sign in is stored on this machine.
# We have not run this script end to end on every platform. It prints and checks, and changes no files.
# Run it in a test shell. Nothing is written to your shell profile.

# 1. Which versions are out, and which do you have?
#    At 5:55 AM ET Thu 5 Nov npm showed @google/gemini-cli latest v0.67.0
#    with preview v0.68.0-preview.0, and @anthropic-ai/claude-code latest v2.1.318.
npm view @google/gemini-cli dist-tags
npm view @anthropic-ai/claude-code dist-tags.latest
if command -v gemini >/dev/null; then gemini --version; else echo "no gemini CLI on PATH"; fi
if command -v claude >/dev/null; then claude --version; else echo "no claude CLI on PATH"; fi

# 2. Optional install only. Pinned at 5:55 AM ET Thu 5 Nov, npm registry channel:
# npm install -g @google/gemini-cli@0.67.0
# npm install -g @anthropic-ai/claude-code@2.1.318
#    Test machine only, npm preview channel, not for production:
# npm install -g @google/gemini-cli@0.68.0-preview.0

# 3. Model discovery from Gemini CLI v0.67.0. Add -o json for structured output.
if command -v gemini >/dev/null; then
  gemini --list-models
fi

# 4. Claude Code v2.1.318 shows loginExpiresAt for a stored claude.ai sign in.
if command -v claude >/dev/null; then
  claude auth status
fi

echo "Done. Nothing was installed or changed on disk."

If gemini --list-models reports an unknown flag, check the version from step 1 before anything else.

Sources:

Recommended AI prompt

Copy this paragraph into ChatGPT, Claude, Gemini, Grok, or whatever AI you use.

Prompt

You are my developer tooling lead for Thursday 5 November 2026. Story 1: Gemini CLI v0.67.0 adds a gemini models command and a flag to list models as JSON with model ids and context limits, stops tool work after a cancel, and bridges the IDE companion into container sandboxes, while v0.68.0-preview.0 stays on test machines only. Story 2: Claude Code v2.1.317 asks before a web or shell tool sends my account email, and v2.1.318 adds loginExpiresAt to claude auth status and credential_expires_at to claude -p JSON, with a stderr warning 10 days before a stored sign in ends. Story 3: Copilot is in public preview inside Xcode's coding assistant starting with Xcode 27.3 beta on macOS Tahoe 26.6 or later, and it joined JetBrains Air the day before. Ask me which of these tools my team uses, whether our scripts hard code Gemini model names, whether we run claude -p on a schedule and whether anyone works in Xcode or Air, then give me a short checklist for each story, and link the official release notes or docs for each step instead of inventing settings or flags.