← All briefs
Matins
27 changes / 4 actionable / 2 deep dives
Claude Code

TL;DR

  • The streaming idle watchdog is now on by default for all providers. It aborts and retries when a response stream goes silent for 5 minutes; set CLAUDE_ENABLE_STREAM_WATCHDOG=0 to disable if you hit false positives. (more below)
  • claude mcp list/get no longer auto-spawn servers from untrusted workspaces. If a repo self-approved MCP servers via a committed .claude/settings.json, they now show ⏸ Pending approval instead of launching silently.
  • --dangerously-skip-permissions was silently falling back to auto mode. It now shows the bypass disclaimer and applies bypass mode to spawned agents as documented.
  • MCP OAuth no longer requests the full scopes_supported catalog when no scope is specified. If you were hitting invalid_scope failures on GitLab self-hosted or enterprise IdPs, this is the fix.

New in 2.1.196

2.1.196 (June 30, 2026)

  • Added support for organization default models; admins set it in the org console, and it shows as "Org default" (or "Role default") in /model when you haven't picked one yourself
  • Added readable default names for sessions at start, making them easier to identify and message
  • Added clickable file attachments in chat; Cmd/Ctrl-click reveals the file in Finder/Explorer
  • Security: claude mcp list/get no longer spawn .mcp.json servers that a repo self-approved via a committed .claude/settings.json; untrusted workspaces show ⏸ Pending approval
  • Fixed waking a background job permanently deleting its conversation and re-running the original prompt when the transcript probe misread a real transcript; the file is now set aside, never deleted
  • Fixed the rate-limit warning flickering off and rate-limit telemetry being over-counted when multiple parallel requests were in flight at the moment a usage limit was hit
  • Fixed duplicate recap lines after a background session's turn: a schema-rejected StructuredOutput attempt no longer renders alongside its retry
  • Fixed PowerShell git diff/git grep, egrep/fgrep, and quoted search patterns containing | being reported as failures when they exit 1, matching Bash behavior
  • Fixed multiple claude agents side panel issues: keyboard focus getting stuck when opening an agent, background jobs losing their subagent types on every open, and sessions showing incorrect status while actively running
  • Fixed claude agents --dangerously-skip-permissions silently falling back to auto mode instead of showing the bypass disclaimer and applying bypass mode to spawned agents
  • Fixed mid-turn crash recovery for Remote sessions; sessions interrupted by a server restart now auto-resume on the next worker
  • Fixed sessions moved with /cd reappearing in the old directory's resume list after a non-graceful exit when the old path contained special characters
  • Fixed claude plugin validate skipping local plugins whose source is "." and stopping after the first error class
  • Fixed Esc Esc at an idle prompt not opening the rewind menu (regression); use Ctrl+C or Ctrl+X Ctrl+K to stop background agents
  • Fixed MCP OAuth requesting the authorization server's full scopes_supported catalog when no scope is specified, causing invalid_scope failures on GitLab self-hosted and other enterprise IdPs
  • Fixed /context showing 0 tokens for all tool groups on Bedrock
  • Fixed /deep-research misreporting verifier failures as "all claims refuted" instead of unverified
  • Fixed plugin dependency version pins not being honored when the marketplace was added as a local folder path backed by a git repo
  • Fixed claude agents session status: completed rows no longer flip between "Done" and "Needs your input", stalled agents are now labeled "Needs attention", and results that mention a PR show a clickable link
  • Fixed voice dictation swallowing spaces and spuriously starting a recording during very fast typing when voice mode is enabled
  • Improved background session reliability: long-running commands and workflows now survive the session's process being stopped, restarted, or updated, including on Windows, where background shells are handed off instead of being killed
  • Improved background agents: workers killed by a daemon restart are now automatically resumed from where they left off the next time the agents view opens
  • Improved /code-review workflow: merged five cleanup finders into one, cutting token usage by roughly 25%
  • Reduced per-frame rendering work in the terminal UI by skipping no-op subtree walks during streaming
  • The streaming idle watchdog is now on by default for all providers. It aborts and retries when a response stream produces no events for 5 minutes. Set CLAUDE_ENABLE_STREAM_WATCHDOG=0 to disable.
  • Remote Control is now disabled when ANTHROPIC_BASE_URL points at a non-Anthropic host, matching the existing behavior under CLAUDE_CODE_USE_BEDROCK/_VERTEX/_FOUNDRY
  • Changed opening the agents view from a foreground session to require a single left-arrow press instead of two, matching the behavior in background sessions

Notes

Streaming watchdog is now opt-out

The watchdog has been evolving across several releases. In 2.1.185, the stream-stall hint was reworked (text changed, trigger moved from 10s to 20s). This release flips the watchdog itself to on-by-default for all providers, meaning stalled streams are now automatically aborted and retried without user intervention. Previously you had to opt in. If you are running against a proxy or provider that legitimately pauses mid-stream (long tool calls, slow inference endpoints), you may see unexpected retries. Set CLAUDE_ENABLE_STREAM_WATCHDOG=0 to revert to the old behavior.

Background agent reliability, the arc continues

This is the third consecutive release tightening background agent lifecycle. 2.1.193 fixed phantom subagent spawns on backgrounding; 2.1.195 fixed agents resurrecting after being stopped and background daemons running unreachable. 2.1.196 goes further: long-running shell commands and workflows now survive process stops, restarts, and updates (including on Windows via shell handoff), and killed workers auto-resume. If you run long background tasks, this release is a meaningful step toward treating background agents as durable rather than fragile.