Unclear The finding does not say what distinguishes 'live' watching from 'durable' watching in practice, beyond the unwatch description.
What
The artifact watch tool's internal checks now branch on a third mode, live, in addition to the existing durable mode (and a default/off state). This changes:
- The permission-ask messaging shown when enabling watch, which now accounts for the
livemode. - The description shown for
unwatchwhen live watching is active: it now says unwatching "stops keeping track of new versions published elsewhere" instead of the old wording about "republish notifications and, in cloud sessions, comment wakes." - The general help text for watch/unwatch/status, reworded from describing "notify" semantics to describing that Claude Code "keeps track of new versions," and clarifying that a republish elsewhere "starts no turn and sends no notification" in live mode.
Why
This distinguishes a lighter-weight "live" tracking mode from the existing "durable" watch mode and corrects the tool's description so it no longer overpromises an active notification, matching the new passive version-tracking behavior.
Documented inapi/agents-and-tools/mcp-tunnels/deploy-helm
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Added since
A small documentation edit on Claude apps gateway for Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry touched a line naming durable after this was published.
| PostgreSQL 11 or later | Backs the device sign-in flow and rate-limit counters. A managed PostgreSQL service works, including the smallest tier; see [which databases are supported](/docs/en/claude-apps-gateway-deploy#postgres). With [spe…claude-apps-gateway see the edit
Confirmed since
Anthropic's documentation has since written up durable, on Supersede older widget instances.
Use [`registerAppTool`](https://modelcontextprotocol.github.io/ext-apps/api/functions/server-helpers.registerAppTool.html) to register the tool, and return the key in `structuredContent` alongside your normal tool output. A per-process cou…connectors/building/mcp-apps/instance-supersession see the edit
Added since
A small documentation edit on How Claude Code uses prompt caching touched a line naming live after this was published.
When you switch [output styles](/docs/en/output-styles) mid-session with [`/output-style`](/docs/en/output-styles#change-your-output-style), `/config`, or the `outputStyle` setting, Claude uses the new style starting with your next message…prompt-caching see the edit
Confirmed since
Anthropic's documentation has since written up live, on Claude Code changelog.
* Fixed four dead documentation URLs (Pricing, Computer Use, Skills, CLI) in the bundled claude-api skill's live-sources tablechangelog see the edit
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubt
The finding does not say what distinguishes 'live' watching from 'durable' watching in practice, beyond the unwatch description.
Anthropic's documentation agrees
Anthropic's documentation has since written up live, on Claude Code changelog.