claudekit / updates / claude-code-2-1-198
[ PATCH · ]

Claude Code 2.1.198

Claude in Chrome is now generally available, background agents that need input or finish now fire the Notification hook, and a new `/dataviz` skill provides chart and dashboard design guidance. Background agents launched from `claude agents` now commit, push, and open a draft PR when they finish worktree code work, and the built-in Explore agent inherits the main session's model (capped at opus) instead of running on haiku. Many bugs were fixed, including brief network drops aborting a turn and background tasks getting stuck on 'Running' after they finish.

Claude Code
Official announcement →

This article is a summary based on official documentation.

What changed

Claude Code 2.1.198 was released on July 1, 2026. It adds new features — Claude in Chrome general availability, background agent notifications, and a /dataviz skill — and improves background agents so they commit, push, and open a draft PR when they finish worktree code work. Fixes span network recovery, background task status, agent teams, and more.

New features

  • Claude in Chrome is now generally available

    Claude in Chrome is now generally available. Having Claude read and act on pages in the browser moves out of preview and into general availability.

  • Background agent notifications

    Background agents launched from claude agents gave no signal when they needed input or finished — you had to open the view to find out. They now fire the Notification hook (agent_needs_input / agent_completed), so you can be notified through hooks.

  • /dataviz skill

    Designing charts and dashboards left color and layout choices up to you. The new /dataviz skill provides chart and dashboard design guidance, with a runnable color-palette validator.

  • Gateway: Claude Platform on AWS upstream provider

    Gateway added Claude Platform on AWS (anthropicAws) as an upstream provider. Model-not-found responses now advance the failover chain to the next provider.

Key improvements

  • Background agents commit, push, and open a draft PR

    Background agents launched from claude agents used to stop and ask what to do after finishing code work in a worktree. They now commit, push, and open a draft PR when they finish instead.

  • Explore agent inherits the main session’s model

    The built-in Explore agent ran on haiku, limiting exploration quality. It now inherits the main session’s model (capped at opus).

  • Subagents and compaction inherit extended thinking config

    Subagents and context compaction did not reflect the session’s extended thinking configuration. They now inherit it, improving output quality on delegated tasks.

  • Removed the /agents wizard

    Creating and managing subagents went through a separate /agents wizard. That wizard is removed; ask Claude to create or manage subagents, or edit .claude/agents/ directly.

  • Highlighting and retry UX

    Syntax highlighting in code blocks, diffs, and file previews was upgraded to highlight.js 11 for better accuracy. On API retries, the error reason is now shown after the second attempt, and a status page link replaces the spinner tip when the API is overloaded.

Bug fixes

Network and connectivity

  • Brief network drops mid-response aborting the turn — transient errors like ECONNRESET now retry with backoff instead of failing.
  • Claude Platform on AWS and Mantle sessions dead-ending with “Please run /login” when the STS token expires — awsAuthRefresh now runs automatically.
  • “no route to host” for local-network hosts in macOS background agent sessions — fixed by declaring Local Network entitlements.
  • Background agents repeatedly showing “Reconnecting…” every ~52 seconds on macOS while the agents view was open — fixed.
  • Excessive background classifier requests when sandboxed processes repeatedly accessed the same network host — fixed.

Background tasks and agents

  • Background tasks in web, desktop, and VS Code task panels getting stuck on “Running” after they finish or after resuming a session — fixed.
  • Agent teams — a teammate that dies on an API error now reports “failed” to the lead, and messaging a stuck teammate wakes it to retry immediately.
  • claude --bg silently creating an unattachable session when combined with --print/-p — the conflicting flags are now rejected up front.
  • The workflow progress view dropping the earliest agents from the list while the phase counter stayed correct in SDK and desktop-app sessions — fixed.
  • Pressing ← inside claude attach <id> exiting to the shell instead of opening the agent view — fixed.

UI and rendering

  • Markdown tables overflowing and wrapping their right border in fullscreen mode — fixed.
  • The /diff panel not refreshing when you switch branches or commit outside the session — fixed.
  • Cmd+click not opening URLs in fullscreen mode in Warp on macOS, and double-click word selection now selects the entire URL including the scheme — fixed.

Commands and config

  • Plan mode not auto-allowing read-only tool calls when a session starts in plan mode — fixed.
  • .claude/rules/ conditional rules not loading when the target file is reached via a symlinked path — fixed.
  • /desktop failing with “Cannot determine working directory” after entering and exiting a worktree — fixed.
  • /branch deriving its default fork name from the compaction summary instead of the first real prompt — fixed.

Notes

  • Removed /agents wizard — ask Claude to create or manage subagents, or edit .claude/agents/ directly.
  • How subagents treat messages — subagents treat messages from the agent that launched them as normal task direction; an agent’s message is still never treated as the user’s approval.
  • Keyboard shortcut hints over SSH from a Mac — hints now show opt/cmd instead of alt/super when connected from a Mac over SSH.