docs: Claude Code v2.1.178-v2.1.204 catch-up — Apps Gateway, desktop-linux, admin API#1046
Conversation
…linux, admin API Manual catch-up fetch. The scheduled pipeline has pushed branches but failed to open PRs since June 14 (claude-code-action@main changed claude_args parsing, splitting 'Bash(gh pr:*)' into invalid rules and silently denying gh pr create), so main is 3 weeks stale. Highlights across the gap: - CHANGELOG: 20 new versions (2.1.178 through 2.1.204) - New doc family: Claude Apps Gateway (config, deploy, GCP, spend limits) - New docs: artifacts, desktop-linux, feature-availability, gateways, llm-gateway protocol/connect/rollout, plugin-relevance - whats-new weekly pages 2026-w23 through w26 - 24 new Admin API docs, per-language SDK references - 1086 modified docs across platform, MCP, GitHub sources Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a06fc7468a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| | Claude Haiku 3.5 ([retired, except on Bedrock and Vertex AI](/docs/en/about-claude/model-deprecations)) | `auto`, `none`<hr />`any`, `tool` | 264 tokens<hr />355 tokens | | ||
| | Model | Tool choice | Tool use system prompt token count | | ||
| | ---------------------------------------------------------------------------------------------------------- | ------------------------------ | ---------------------------------- | | ||
| | Claude Opus 4.8 | `auto`, `none`***`any`, `tool` | 290 tokens***410 tokens | |
There was a problem hiding this comment.
For every row in this pricing table, the previous <hr /> separator between the two tool-choice/token-count cases was replaced with ***. Markdown treats *** as emphasis markup rather than a visible row separator, so the rendered docs collapse the auto/none and any/tool cases and make it unclear which token count applies to which tool choice, which can mislead users estimating tool-use costs.
Useful? React with 👍 / 👎.
Manual catch-up (day shift, human-triggered)
WHY THIS MATTERS: main has been stale since June 14 while Anthropic shipped 20 Claude Code releases. This PR restores the archive to current.
WHAT CHANGED:
claude-apps-gateway*.md— config, deploy, GCP, spend limits)artifacts,desktop-linux,feature-availability,gateways,llm-gateway-{connect,protocol,rollout},plugin-relevance, whats-new weekly pages w23–w26Why the pipeline went dark (root cause)
The scheduled workflow tracks
anthropics/claude-code-action@main. Around June 14–18 the action'sclaude_argsparsing changed:--allowedTools Bash(gh pr:*),…now splits on the space intoBash(gh+pr:*)— two invalid rules. Night-shift Claude could still commit and push (Bash(git:*)survived) but everygh pr createwas silently denied. Result: 31 orphaneddocs/claude-code-v2.1.*branches (June 18 → today), Bark notifications with no PRs, and green runs that did nothing.Separately, the fetcher fails on 731/3,638 URLs (157 = new platform
/api/terraform/*pages with no.mdvariant; 574 = r.jina.ai rate limiting on blog/support) and discards failure records from.metadata.json, so this was invisible.Pipeline fix lands in a follow-up PR. Orphan branches can be deleted once this merges.
🤖 Generated with Claude Code