# Claude Code v2.1.229 > Claude Code v2.1.229, released 12 Aug 2026 (2026-08-12). 205 entries read out of the shipped bundle. Unofficial, and not affiliated with Anthropic. Web version: https://changelogs.core-directive.com/v/2.1.229 Thirty-nine entries are present but not switched on. A large `/design` skill drafts multi-artboard canvases published as an editable Artifact, but it needs the `tengu_ethereal_nova` gate, which falls back to false; a slides artifact template with a thumbnail rail, comment threads and a Present mode is registered as a command whose enable check returns false outright. A built-in `/commit` command and a new `pr` skill that opens pull requests through the gh CLI both wait on the `tengu_cinder_relay` gate, which the build compiles as false and the server decides. Rate-limit handling can queue an automatic continue at reset and nudge Claude to wrap up when five-hour usage runs low, both behind `tengu_vellum_anchor`, also off. A `ReadNotifications` tool and the queued notifications that feed it need `CLAUDE_CODE_REMOTE` set and a non-interactive session, and `CLAUDE_CODE_ARTIFACT_VIEWER_BASE_URL` is recognised but read by a resolver with an empty body. Ninety-four entries are usable now. `CLAUDE_CODE_TURN_UPDATES` turns on a narration style where Claude says what it is about to do, posts progress and writes a standalone final message. `/teleport` gained "Continue this session in the cloud", which requires Remote Control, a claude.ai login and a clean git tree with nothing unpushed. Plugins can now be installed from a `command` source, a shell command that prints a plugin directory, with a `timeout` capped at 600 seconds and a `mode` of copy or link; administrators can block it with the managed setting `disableCommandPluginSources`, and a repository can set `disableAllHooks: true`. The `[1m]` long-context variants are no longer hidden from subscription accounts pointed at a custom `ANTHROPIC_BASE_URL`, and self-hosted runners on Windows now demand `--base-dir` instead of falling back to a POSIX path. MCP sign-in now redirects to `http://127.0.0.1` rather than `localhost`, fixing machines where localhost resolves to an unlistened IPv6 address, and a stale saved OAuth client is repaired in place instead of forcing a fresh registration. A failed sandbox initialization clears its state rather than caching the rejected promise, so a later attempt can succeed. Messages injected into a session programmatically are no longer counted as verified human turns, and tool inputs are filtered to a fixed set of fields before being recorded in history. The bundled `code-walkthrough` and `pr-explainer` skills are gone, and `PowerShell(git checkout -b *)` is no longer pre-allowed. The billing attribution header builder now takes an options argument, and one first-party, non-socket path passes `ignoreEnvOptOut` so the header is sent even when `CLAUDE_CODE_ATTRIBUTION_HEADER` opts out. The auto-mode classifier, the two-stage classifier and the critique query do the same; the block carries version, entrypoint and workload metadata rather than conversation content. Plugins installed from a command re-resolve in the background once per session, subject to non-essential network traffic being allowed and a remote flag that defaults on. ## What probably matters to you Anything you can use today, anything that visibly changes, and anything worth poking at. One line each, open for detail. ### "Turn updates" narration style for Claude's replies Claude can narrate what it's about to do and post progress updates; set CLAUDE_CODE_TURN_UPDATES. **What** Claude can be told to say what it is about to do before starting, post brief progress updates while working, and write a final message that stands alone. Set `CLAUDE_CODE_TURN_UPDATES` to turn it on or off yourself; otherwise it is decided by a per-model capability check, then by a capability map the server sends with your account data. It replaces the usual communication guidance in the system prompt rather than adding to it. **Details** - Resolution order: the environment variable wins if set (it accepts true/false/unset), then the model capability check, then the server-supplied map. - The env var is new in this build. - When the value comes from the server-supplied map, a `tengu_model_capability_from_client_data` telemetry event is emitted. **Usage** `CLAUDE_CODE_TURN_UPDATES=1 claude` **Evidence** `Before you start, say in a line what you're about to do; brief updates while you work help the user follow along. Your final message should stand on its own for a reader catching up fresh.` - Area: Output Style - Names: `CLAUDE_CODE_TURN_UPDATES` - Tier: Use it now - Useful: 5/5 - Signal: 5/5 ### A slides artifact template exists but its command is switched off A presentation-deck artifact with slide editing, comments and Present mode is built but unreachable. **What** A third artifact family joins doc and sheet: a presentation deck, with a full skill body and a published-deck template offering a thumbnail rail, in-place slide editing, comment threads kept in a hidden storage block, and a Present mode. It is registered as a user command described as "Publish a presentation deck Artifact", but the registration's isEnabled function returns false in this build, so the command cannot be reached. **Details** - Present mode: arrow keys step, N toggles speaker notes, B blanks the screen, T runs a timer, digits jump to a slide, Esc exits. - Nothing in the build makes isEnabled return anything other than false. **Evidence** `Publish a presentation deck Artifact` - Area: Artifacts - Tier: Not switched on - Useful: 3/5 - Signal: 5/5 - Present in the build but not switched on ### New /design skill drafts multi-artboard canvases as an Artifact, gated off A /design skill drafting multi-artboard canvases as editable Artifacts is bundled but switched off. **What** A large new bundled skill drafts designs as `.dc.html` artboards laid out on a single pan-and-zoom canvas and publishes them as an Artifact, with an editor built into the published page. Saving republishes for everyone only where the artifact grants the self-write capability; otherwise the page is view plus PNG and PDF export. It registers a `design` slash command. Off in this build: it needs `tengu_ethereal_nova`, which falls back to false, and an Artifact tool that advertises capabilities. **Details** - ships as a skill document plus a canvas seeding script and a precompiled editor page payload **Evidence** `Draft a design on a canvas Artifact — editable where saving is enabled (Claude Design preview)` - Flag `tengu_ethereal_nova`: On for this account, and not off by default (read for one account on one subscription tier against v2.1.229; this account: on, anonymous baseline: on, compiled default: on) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release. - Area: Artifacts - Names: `/design` - Tier: Not switched on - Useful: 3/5 - Signal: 5/5 - Present in the build but not switched on ### Remote sessions can receive queued notifications and read them with a new tool Remote sessions can be handed webhook, schedule and cross-session events mid-turn and read them with a new tool. **What** Sessions running in remote mode can now be handed out-of-band events while a turn is in flight (GitHub webhook activity, a scheduled trigger firing, a message from another Claude session). They are buffered and Claude is nudged to call a new tool, ReadNotifications, until it reports 0 remaining. This needs remote mode, set through `CLAUDE_CODE_REMOTE` plus an internal remote check; payloads arriving otherwise are dropped. **Details** - Drained notification bodies are wrapped in an explicit warning that they are relayed external content and may imitate the surrounding delimiters. - Oversized content is refused without acknowledgement, and the buffer has a fixed size cap. - Telemetry lands under `ccr_queued_notifications` with reasons `malformed_payload`, `not_remote`, `content_too_large`, `buffer_full` and `nudge_ignored`. **Evidence** `queued_notification outside remote mode` - Area: Remote Control - Names: `CLAUDE_CODE_REMOTE` - Tier: Not switched on - Useful: 3/5 - Signal: 5/5 - Present in the build but not switched on ### /teleport can send the current session to the cloud /teleport can hand your current session to the cloud if your git tree is clean and pushed. **What** The teleport menu gained "Continue this session in the cloud", which hands the current session to a cloud session and then either reconnects Remote Control to it or disconnects. It refuses unless Remote Control is active, you are logged in to claude.ai, and you are in a git repo with a clean tree and nothing unpushed. Available where /teleport already is, which needs the `allow_remote_sessions` remote setting. **Details** - calls a new `move-to-cloud` endpoint on the session after the precondition checks pass - the command description changed to cover both directions of travel - emits `tengu_teleport_to_cloud` and `teleport_to_cloud` telemetry **Evidence** `Send this session to the cloud, or resume one from claude.ai` - Area: Remote Control - Names: `/teleport`, `allow_remote_sessions` - Tier: Use it now - Useful: 4/5 - Signal: 4/5 ### ReadNotifications tool, reachable only in remote non-interactive runs A new tool lets Claude pull queued notifications, but only in remote non-interactive runs. **What** A new read-only tool called ReadNotifications lets the model pull notifications the server has queued for a session (GitHub activity, scheduled triggers and messages from other agents), oldest first, marking them delivered and reporting how many are left. It is only offered when the `CLAUDE_CODE_REMOTE` environment variable is set and the session is non-interactive, so it does not appear in an ordinary terminal session. **Details** - Each notification carries an id, an origin, a queue timestamp and its content, plus a count of those still waiting. - Subagents are refused: "ReadNotifications can only be called from the main conversation — subagents must not drain the session notification buffer." - This replaces injecting webhook and trigger activity into the conversation as if you had typed it. **Evidence** `ReadNotifications can only be called from the main conversation — subagents must not drain the session notification buffer.`, `Read the notifications queued for this session` - Area: Sessions - Names: `CLAUDE_CODE_REMOTE` - Tier: Not switched on - Useful: 2/5 - Signal: 5/5 - Present in the build but not switched on ### Windows now has its own gate for Harbor cross-session messaging Setting CLAUDE_CODE_HARBOR_KITE forces cross-session messaging on, and Windows can now be allowed through. **What** The check for Harbor cross-session messaging was rewritten. Setting `CLAUDE_CODE_HARBOR_KITE` now forces it on regardless of platform. Windows, which previously always returned false, is allowed through when the separate `tengu_harbor_kite_win` gate is on. The general gate is consulted last. Both gates fall back to off in this build, so unless the environment variable is set, whether it runs is decided by remote config. **Evidence** `tengu_harbor_kite_win` - Flag `tengu_harbor_kite_win`: On for this account, and not off by default (read for one account on one subscription tier against v2.1.229; this account: on, anonymous baseline: on, compiled default: on) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release. - Area: Cross-Session Messaging - Names: `CLAUDE_CODE_HARBOR_KITE` - Tier: Use it now - Useful: 3/5 - Signal: 4/5 ### Artifact watches can wake on comments as well as republishes Artifact watches can now wake you on comments addressed to Claude, not just republishes. **What** Subscribing to an artifact now returns which events it covers, and the status line says whether you will be woken on republishes only or on republishes plus comments addressed to Claude. **Details** - A path that arms these wake subscriptions automatically is present but sits behind a build-time constant that is `false` in this build, so it never runs. **Evidence** `watching for republishes and to-Claude comments` - Area: Artifacts - Tier: You'll notice - Useful: 3/5 - Signal: 4/5 ### A built-in /commit command, off unless a server gate turns it on A built-in commit command with a strict git workflow exists but is off unless a server enables it. **What** A command named "commit" is registered that gathers git status, the diff against HEAD, the branch and the last 10 commits, then follows a fixed workflow: never amend after a failed pre-commit hook, never use `--no-verify`, prefer naming files over `git add -A`, and never push unless asked. Its tool access is limited to git add, status and commit. It appears only when the `tengu_cinder_relay` gate resolves true; the value compiled into the build is false, so whether it shows up is decided by server config. **Details** - It is separate from the existing commit-push-PR flow. **Evidence** `Create a git commit. Use whenever you are about to create a commit`, `tengu_cinder_relay` - Area: Slash Commands - Names: `/commit` - Tier: Not switched on - Useful: 3/5 - Signal: 4/5 - Present in the build but not switched on ### New pr skill opens pull requests, replacing pr-explainer A pr command that opens GitHub pull requests is bundled but gated off; pr-explainer is gone. **What** A new `pr` command gathers git context and opens a GitHub pull request through a fixed gh CLI workflow, with a fixed allowed-tools list covering `gh pr create --title * --body *` and `git push -u origin *` and a heredoc body format. Off in this build: it needs `tengu_cinder_relay`, which falls back to false, behind the commit-skill rollout gate. The old `pr-explainer` command that generated a shareable PR walkthrough artifact is gone, and PR review skills no longer point at it. **Evidence** `Create a GitHub pull request. Use whenever you are about to open a PR` - Area: Slash Commands - Names: `/pr` - Tier: Not switched on - Useful: 3/5 - Signal: 4/5 - Present in the build but not switched on ### Nudge to wrap up when your 5-hour usage is nearly spent Claude could be told to wrap up and list remaining work when your usage is nearly spent. **What** When five-hour usage crosses a threshold, and you are not using or allowed extra overage, Claude Code arms a one-shot hint that tells the model to finish the current step, list at most 3 bullets of the most impactful remaining work, and not start subagents or long-running work. Off in this build: `tengu_vellum_anchor` falls back to false. Reports as `usage_limit_near_wrapup`. **Details** - fires at most once per session, injected as a meta message rather than shown as UI text **Evidence** `[Usage limit approaching. Checkpoint now: finish the current step, then list up to 3 short bullets of the most impactful remaining work. Don't start subagents or long-running work.]` - Flag `tengu_vellum_anchor`: Off in both readings (read for one account on one subscription tier against v2.1.229; this account: off, anonymous baseline: off, compiled default: on) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release. - Area: Rate Limits - Tier: Not switched on - Useful: 3/5 - Signal: 4/5 - Present in the build but not switched on ### Rate-limit banner can queue an automatic continue at reset Hitting a five-hour limit could queue your turn to resume automatically at reset, currently off. **What** When you hit a five-hour limit and the reset time is known, Claude Code can queue your turn to resume after the reset and show a hint line ending in " to continue after reset", or prefill `continue` into the prompt. Choosing "adjust" or "upgrade" from the rate-limit or spend-limit menu cancels the queued continue; cancelling or choosing "wait" prefills it instead. The whole path is behind the `tengu_vellum_anchor` flag, which falls back to off, so nothing in this build turns it on locally and whether it appears is decided by server config. **Details** - Events: `tengu_rl_checkpoint_copy_shown` when the hint is shown, `tengu_rl_checkpoint_auto_continue_queued` when the continue is armed, and a third when the upsell path cancels it - Hint text is assembled from a key chord plus the fixed trailing phrase **Evidence** `tengu_rl_checkpoint_auto_continue_queued`, `Press ` - Flag `tengu_vellum_anchor`: Off in both readings (read for one account on one subscription tier against v2.1.229; this account: off, anonymous baseline: off, compiled default: on) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release. - Area: Rate Limits - Tier: Not switched on - Useful: 3/5 - Signal: 4/5 - Present in the build but not switched on ### Wrap-up hint when you are close to the usage limit Past 95% of five-hour usage Claude could be told to checkpoint and list remaining work, off here. **What** When five-hour usage crosses 95% and the account is not in overage, Claude gets a message telling it to checkpoint: finish the current step, list up to three bullets of the most impactful remaining work, and start no subagents or long-running work. A `usage_limit_near_wrapup` event is recorded. The whole path is behind the `tengu_vellum_anchor` gate, which falls back to off, so with no remote config value nothing is injected. **Details** - The exact injected text is "[Usage limit approaching. Checkpoint now: finish the current step, then list up to 3 short bullets of the most impactful remaining work. Don't start subagents or long-running work.]". - None of this machinery existed in 2.1.228. **Evidence** `tengu_vellum_anchor` - Flag `tengu_vellum_anchor`: Off in both readings (read for one account on one subscription tier against v2.1.229; this account: off, anonymous baseline: off, compiled default: on) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release. - Area: Rate Limits - Tier: Not switched on - Useful: 3/5 - Signal: 4/5 - Present in the build but not switched on ### Attribution header can override the environment opt-out One first-party request path sends the attribution header even if you opted out via env var. **What** The builder for the attribution header now takes an options argument, and one specific first-party, non-unix-socket path passes `ignoreEnvOptOut`, which makes the header be sent even when `CLAUDE_CODE_ATTRIBUTION_HEADER` is set to opt out. **Evidence** `ignoreEnvOptOut` - Area: Telemetry - Names: `CLAUDE_CODE_ATTRIBUTION_HEADER` - Tier: Under the hood - Useful: 3/5 - Signal: 4/5 ### Auto-mode side queries always send the billing attribution header Auto-mode's side queries send billing attribution metadata even if you set the opt-out variable. **What** The auto-mode classifier, the two-stage classifier and the critique query now force the billing attribution system block to be sent even when the opt-out environment variable is set. It carries version, entrypoint and workload metadata, not conversation content, and is only sent on the first-party API and never over a unix socket. **Evidence** `forceAttributionHeader` - Area: Telemetry - Names: `CLAUDE_CODE_ATTRIBUTION_HEADER` - Tier: Under the hood - Useful: 3/5 - Signal: 4/5 ### /prototype now picks a fidelity and can build on an app you already have /prototype picks a fidelity itself, can build on your existing app, and hands off a scoped brief. **What** The prototype skill names three fidelities, sketch, clickable and wired, defaults to clickable, and states its pick instead of asking you. If you explicitly ask, it can prototype a feature on top of an existing app, either as a screenshot overlay or a shell built from your repo source. A new hand-off step converts the finished prototype into an approved brief split into must-have, nice-to-have and cut. **Evidence** `Every prototype is built at one of three fidelities, and naming it sets` - Area: Skills - Names: `/prototype` - Tier: Use it now - Useful: 4/5 - Signal: 3/5 ### 1M-context models no longer suppressed on custom endpoints Subscription users on a custom base URL now see the 1M-context model variants again. **What** The `[1m]` long-context model variants were hidden from subscription accounts whose credit state disallowed extra usage. That credit check now applies only when you are talking to Anthropic's own endpoints or a local socket, so subscription users pointed at a custom `ANTHROPIC_BASE_URL` see the 1M variants offered. Setting `CLAUDE_CODE_DISABLE_1M_CONTEXT` still hides them everywhere. **Details** - The first-party test is satisfied by `ANTHROPIC_UNIX_SOCKET` being set, or by the base URL being an Anthropic one. - On first-party endpoints behaviour is unchanged: cached extra-usage state still decides. **Evidence** `CLAUDE_CODE_DISABLE_1M_CONTEXT` - Area: Models - Names: `ANTHROPIC_BASE_URL`, `CLAUDE_CODE_DISABLE_1M_CONTEXT`, `[1m]` - Tier: Use it now - Useful: 4/5 - Signal: 3/5 ### Directory sync can be turned off with an environment variable You can turn off background directory syncing by setting CLAUDE_CODE_DISABLE_DIR_SYNC. **What** The background worker that syncs directory contents now starts only when a decision helper allows it, and `CLAUDE_CODE_DISABLE_DIR_SYNC` feeds that decision alongside the entrypoint being used. When it is skipped for that reason, a telemetry line records it. **Evidence** `dir_sync_worker_disabled` - Area: File Sync - Names: `CLAUDE_CODE_DISABLE_DIR_SYNC` - Tier: Use it now - Useful: 4/5 - Signal: 3/5 ### Suggested permission rules for commit and PR commands narrowed Approving commit and PR permissions now allows only specific flags, not any git or gh arguments. **What** The built-in allow-rule suggestions are now argument-specific: `git commit -m *`, `gh pr create --title * --body *` and `gh pr edit --title * --body *`. Approving one no longer blanket-allows arbitrary commit or pull-request flags. **Evidence** `gh pr create --title * --body *` - Area: Permissions - Tier: You'll notice - Useful: 4/5 - Signal: 3/5 ### OpenTelemetry can be sent through Anthropic's first-party gateway Telemetry can route through Anthropic's own gateway, but only under a narrow set of conditions. **What** Telemetry export now decides per signal (metrics, logs, traces) whether to send OTLP data to a gateway endpoint derived from `ANTHROPIC_BASE_URL`, authenticated with a `Bearer` header from `ANTHROPIC_AUTH_TOKEN`. It only does this when `CLAUDE_CODE_PROVIDER_MANAGED_BY_HOST` is set, the auth mode is first-party/gateway, the token is scoped to the gateway audience, you have set no OTLP headers of your own, and your OTLP endpoint already matches the derived URL. Any other combination sends nothing. **Details** - The check runs separately for each signal, so metrics can qualify while traces do not. - User-set `OTEL_EXPORTER_OTLP_*` headers disqualify the signal outright, so your own collector configuration is never overridden. - Falls back to returning "do not send" rather than erroring. **Evidence** `claude-gateway` - Area: Telemetry - Names: `ANTHROPIC_BASE_URL`, `ANTHROPIC_AUTH_TOKEN`, `CLAUDE_CODE_PROVIDER_MANAGED_BY_HOST` - Tier: Use it now - Useful: 2/5 - Signal: 4/5 ### Artifact comment auto-replies report why they did not reply Automatic replies to artifact comments now record why they skipped a thread. **What** The pipeline that reacts to comments on published artifacts now records a status for each comment thread rather than just succeeding or vanishing: declined with a reason such as blocked, auto_reply_off, plan_mode, hourly_cap, loop_guard or denied_by_rule, or failed with compose_failed or post_failed. Threads seen for the first time can be deferred instead of dropped. Nothing in the build resolves whether the surrounding auto-reply feature is on for a given account. **Evidence** `"deferred_first_sight"` - Area: Artifacts - Tier: Not switched on - Useful: 2/5 - Signal: 4/5 - Present in the build but not switched on ### Artifact database gains a private per-viewer area Artifact databases can give each viewer a private area, documented only when an internal flag is on. **What** Artifact database guidance documents an exception to shared rows: each viewer's subtree under the `data/users/` prefix is private to that viewer, and the literal path segment `me` stands for the current viewer's id, the same one the page gets from `claude.user.id()`. It requires the published version to declare the `user` capability alongside `db`. The paragraph is only included in the tool description when its internal flag is set. **Evidence** `data/users/me` - Area: Artifacts - Tier: Not switched on - Useful: 2/5 - Signal: 4/5 - Present in the build but not switched on ### Artifact publish can demand a session context, switched off here Publishing an artifact could carry session context so a later session resumes it, but it's disabled. **What** Publishing an artifact gained an optional `session_context` parameter, an error when it is missing on a first publish, and capture of the prompt that triggered the publish (truncated with a marker) plus the model into the published metadata. It all sits behind a latch that is initialized to false in this build, so the parameter is never offered and the requirement never fires. **Details** - Error case is `session_context_required` on the first publish only - Latch has no code path in this build that sets it true **Evidence** `session_context_required` - Area: Artifacts - Tier: Not switched on - Useful: 2/5 - Signal: 4/5 - Present in the build but not switched on ### Artifact watches can wake on comments, and say why they did not Watching an artifact could wake on new comments, not just republishes, but it's server-gated off. **What** In remote sessions, watching a published artifact can now register a wake that fires on new comments as well as republishes, and the tool text was rewritten to match. The whole durable wake path is behind a remote-config flag whose built-in fallback is false, so it is off unless the server turns it on, and it also requires the remote-session mode. **Details** - Watch results can carry a skip reason, shown as "durable wake subscriptions are not enabled for this session" or "no wake-webhook minter is available (the session's own MCP mount was not found)". - The gate is `tengu_scalable_quiche`, alongside `CLAUDE_CODE_REMOTE`. **Evidence** `This remote session's durable wake rail was skipped as well` - Flag `tengu_scalable_quiche`: Off in both readings (read for one account on one subscription tier against v2.1.229; this account: off, anonymous baseline: off, compiled default: on) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release. - Area: Artifacts - Tier: Not switched on - Useful: 2/5 - Signal: 4/5 - Present in the build but not switched on ### Custom artifact viewer host, not yet resolvable An env var for a custom artifact viewer host is recognised but currently ignored. **What** `CLAUDE_CODE_ARTIFACT_VIEWER_BASE_URL` is a new recognised environment variable, and artifact URL parsing and building now ask a resolver for the viewer host instead of assuming claude.ai or the staging host. In this build that resolver has an empty body and returns nothing, so the variable is never read and URLs still resolve to the built-in hosts. **Evidence** `CLAUDE_CODE_ARTIFACT_VIEWER_BASE_URL` - Area: Artifacts - Names: `CLAUDE_CODE_ARTIFACT_VIEWER_BASE_URL` - Tier: Not switched on - Useful: 2/5 - Signal: 4/5 - Present in the build but not switched on ### PR writing guidance and PR templates are present but switched off Plain-language PR writing rules and repo PR template fetching exist but reach no prompt. **What** Helpers were added that would inject plain-language PR writing rules ("Short beats complete"), fetch the repo's PR template out of git, and warn that the template is untrusted content. All of it sits behind a helper that unconditionally returns false in this build, so none of it reaches any prompt. **Details** - The template fetch uses `git show HEAD: