Docs, Slides and Design Leave Beta on Every Plan
Claude Docs, Slides and Design are now generally available on all plans, including Free. New features include customer-managed encryption keys for artifacts, team co-editing, PowerPoint and PDF export, and mobile editing. Dashboards (paid plans) and Motion (Team and Enterprise) remain in beta.
Enterprise admins: Docs, Slides and Design turn on by default October 15. Dashboards and Motion are off until enabled under Organization settings. The standalone claude.ai/design stays available until December 14, and admins can migrate design systems from the Artifacts page.
Haiku 5.5: Cheap Until You Cross 100K Tokens
Claude Haiku 5.5 (claude-haiku-5-5) is now the default Haiku on the API with a 1M context window. Short prompts cost $0.10 input and $0.50 output per million tokens, about a tenth of Haiku 4.5’s $1 and $5.
The catch: prompts over 100K tokens jump to $0.50 and $2.50, roughly 5x Haiku 4.5. Long-context workloads should do the math before switching.
Max and Team Plans Get Monthly API Credits
Max and Team plans now include monthly API credits for running apps and agents on the Claude Platform. Claim them by linking a Claude Console organization in your billing settings.
Claude Code Closes a Run of Permission Bypasses
Fixes across 2.1.289 to 2.1.296 cover several ways to slip past rules. Bash commands with an environment-variable prefix, like TZ=... rm -rf build, skipped deny and ask rules when the sandbox auto-allowed commands. Read deny rules did not apply to files reached through a symlink, and PreToolUse hook approvals and auto mode skipped the prompt for file reads on UNC network paths.
On macOS and Windows, a notebook or PDF read could also return a file outside what was approved if a link was swapped mid-read. The digest’s advice: be on 2.1.296 or later.
Hooks Can Now Fail Closed
Claude Code 2.1.295 adds onFailure: block for command and HTTP hooks. A hook that fails to start, times out or exits unexpectedly now blocks the action instead of letting it through. Version 2.1.294 also fixed prompt and agent hooks written as instructions that sometimes allowed actions they should have blocked.
Also in 2.1.295: terminal support for the OSC 7501 status protocol, so terminals can show whether Claude Code is working, waiting or done. Version 2.1.296 adds the Read tool’s allow_large option and raises the MCP tool description limit to 4,096 characters.
This Week’s Incidents: Two Blips, No Core Outage
On October 5, Anthropic’s status page logged about 30 minutes of elevated errors on Claude Mythos 5.1 and Claude Fable 5.1, touching Claude Code, the API, claude.ai and Cowork. It was resolved the same morning.
On October 7, platform.claude.com showed elevated errors loading usage data for about eight hours, and the Admin API usage report was affected. Core functionality and API requests were not. The latest roughly 30 minutes of usage may lag while the fix finishes.
The Boring Releases Are the Important Ones
Nothing flashy today, and that is the point. Env-var prefixes, symlinks and silent hook failures are exactly how an agent, or a sloppy config, walks around a guardrail. A hook that fails closed is worth more than another benchmark bump.
Pair that with a Haiku 5.5 price list that rewards short prompts and a free-for-all Docs, Slides and Design launch, and the week’s pattern is clear: Anthropic is tightening the plumbing while widening the audience. Patch your CLI before Monday.