Claude Code 2.1.81 adds a bare automation mode that skips hooks, LSP, plugin sync, and skill scans, plus a channels relay for phone approvals. It matters for safer scripted runs and lower-context tool calls, especially in multi-session setups.
--bare flag for scripted -p calls, and the changelog thread says it skips hooks, LSP, plugin sync, and skill directory walks for faster, more deterministic automation.--channels permission relay so channel servers with the right capability can forward tool approval prompts to your phone.400The headline change is --bare, which Anthropic describes in the changelog thread as a mode for scripted -p calls that skips hooks, LSP startup, plugin sync, and skill directory walks. That makes the CLI behave more like a stripped-down automation runtime than an interactive coding environment. The same note says --bare requires ANTHROPIC_API_KEY or an apiKeyHelper passed via settings, with OAuth and keychain auth disabled and “auto-memory fully disabled.”
That combination matters because it narrows side effects during non-interactive runs. In the same release family, the tracked CLI surface in surface diff also shows a new critique command, a new MCP_OAUTH_CLIENT_METADATA_URL environment variable, and new config keys including showClearContextOnPlanAccept.
The second feature is --channels permission relay. Anthropic’s changelog thread says channel servers that declare the permission capability can send tool approval prompts to a phone, extending the approval loop outside the terminal.
The system prompt changes are small but operationally relevant. According to the prompt update, memory access triggers are now consolidated to cases where prior work is relevant, and there is a hard rule that if a user asks Claude to ignore memory, it must “not cite, compare against, or mention memory” and should answer as if none exists. The same update also switches Read-tool guidance from whole-file reads toward targeted reads of known sections, especially for large files.
The rest of 2.1.81 is mostly bug fixing. Anthropic says it fixed repeated re-authentication when multiple Claude Code sessions were open, voice-mode failures that surfaced as a generic “check your network” message, and a bug where the structured-outputs beta header could trigger 400 errors on proxy gateways forwarding to Vertex or Bedrock changelog thread. The release also fixes a race where background agent task output could hang indefinitely, restores Bash PATH inheritance on Windows with Git Bash in VS Code, and adds MCP OAuth support for Client ID Metadata Document (CIMD / SEP-991) for servers without dynamic client registration release thread.
Claude can now drive macOS apps, browser tabs, the keyboard, and the mouse from Claude Cowork and Claude Code, with permission prompts when it needs direct screen access. That makes legacy desktop workflows automatable, and Anthropic is pairing the push with more background-task support for longer agent loops.
releaseOpenClaw shipped version 2026.3.22 with ClawHub, OpenShell plus SSH sandboxes, side-question flows, and more search and model options, then followed with a 2026.3.23 patch. Teams get a broader plugin surface, but should patch quickly and review plugin trust boundaries as the ecosystem grows.
releaseCursor shipped Instant Grep, a local regex index built from n-grams, inverted indexes, and Bloom filters that drops large-repo searches from seconds to milliseconds. Faster candidate retrieval shortens the coding-agent loop, especially when ripgrep-style scans become the bottleneck.
breakingChatGPT now saves uploaded and generated files into an account-level Library that can be reused across conversations from the web sidebar or recent-files picker. It removes repetitive re-uploading and makes past PDFs, spreadsheets, and images part of a persistent working context.
breakingEpoch AI says GPT-5.4 Pro elicited a publishable solution to one 2019 conjecture in its FrontierMath Open Problems set, with a formal writeup planned. Treat it as an early milestone worth reproducing, not blanket evidence that frontier models can already automate math research.
Claude Code 2.1.81 system prompt updates Notable changes: 1) Claude’s memory policy changes: memory access triggers are consolidated (“when relevant or prior work is referenced”), and a new hard rule says if the user asks to ignore memory, Claude must not cite, compare against, Show more
Claude Code 2.1.81 has been released. 27 CLI changes, 2 system prompt changes Highlights: • Added --bare flag: skips hooks, LSP, plugin sync and skill scans for faster, deterministic automation • Assistant honors 'no memory' requests: ignores stored memory and reports none, Show more
Claude Code CLI 2.1.81 changelog: New features: • Added --bare flag for scripted -p calls — skips hooks, LSP, plugin sync, and skill directory walks; requires ANTHROPIC_API_KEY or an apiKeyHelper via --settings (OAuth and keychain auth disabled); auto-memory fully disabled • Show more
