Top 5
CapitalCyera Acquires Oasis Security for $1B as Non-Human Identity Becomes a Platform WarPolicyDario Amodei Breaks His Silence. Anthropic Has Never Called for an Open-Weight Ban.Agents & ToolsAnthropic never shipped a noindex tag. Hundreds of Claude shared chats landed on Google.Policy90 Companies Now. None of Them Had a Choice.PolicyThe Open-Weight Letter Is a White Flag Disguised as a ManifestoCapitalCyera Acquires Oasis Security for $1B as Non-Human Identity Becomes a Platform WarPolicyDario Amodei Breaks His Silence. Anthropic Has Never Called for an Open-Weight Ban.Agents & ToolsAnthropic never shipped a noindex tag. Hundreds of Claude shared chats landed on Google.Policy90 Companies Now. None of Them Had a Choice.PolicyThe Open-Weight Letter Is a White Flag Disguised as a Manifesto

The AI Sift is part of you-do-nothing

The AI Sift
Subscribe
← Back to The Latest
Agents & Tools2w ago

Claude Code CLI 2.1.209: bash tool, agent launcher, /model fix

Claude Code shipped a 3-item CLI update. Bash tool and agent launcher land as first-class tools for verification and sub-agent delegation. The /model fix restores dialogs in background agent sessions. 310 words, daily tier.

Context from: Github

The decision it puts on your desk

Update to 2.1.209 by end of day. Wire the bash tool into one existing verification workflow before end of week. If you have been deferring the sub-agents pattern, the agent launcher makes it runtime-native now. Adopt the pattern on one multi-step workflow by end of sprint.

Claude Code shipped a 3-item CLI update in version 2.1.209. Two of the three items are unblockers for agentic workflows. The third fixes a UX bug introduced in a prior release.

A bash tool is now first-class. Prompts can run shell commands and get the output back, which means the loop can call its own verifier. Test suites, linters, builds, custom gates: anything that exits with a status code is now reachable from inside a Claude Code session without custom wiring.

An agent-launcher tool is the second item. Prompts can spawn new agents for complex multi-step tasks, chain them, delegate. The sub-agents building block we wrote about in the Karpathy Loop story just stopped being a manual architecture decision. The runtime ships it.

The /model fix is the quiet one. Background agent sessions no longer block /model and other dialogs, restoring the ability to switch models mid-session while an agent runs. The fix is a revert of an overly broad guard that landed in the prior release. If you have been stuck watching a spinner when you tried to change models in a background session, that is gone.

Run claude --version and confirm 2.1.209 by end of day. Test the bash tool on one verification task this week (a test suite, a build, a lint). If you have been blocked by /model hangs in background sessions, this is the fix.

Source

https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md#21209

← Previous

14.8k stars, 2.2k forks, 16 free LLM providers. FreeLLMAPI just productized the free-tier stack.

Next →

Claude Code artifacts: 3 features turn chat output into team infrastructure

The Morning Recap

We sift the day. You make the calls.

The decision on every story, in your inbox before your first call.