🧭 Claude Cowork Suite Reaches General Availability — Whiteboards, Sheets, and Data Open to All Paid Plans from Today
Three days after the initial announcement, Anthropic opens its Cowork productivity suite to general availability on November 14. Claude Whiteboards, Claude Sheets, and Claude Data (formerly Claude Dashboards) are now accessible on paid Team and Enterprise plans, with admin controls governing which products are on by default. Claude Data turns on for all eligible organisations immediately; Whiteboards and Sheets require an explicit admin opt-in until November 20, after which they will be default-on enterprise-wide.
What's now GA
- Claude Whiteboards: Real-time collaborative canvas. Claude generates shapes, swimlanes, connectors, and annotation blocks in response to spoken prompts during calls or typed instructions. Exports as PDF or SVG. Available in claude.ai web; mobile support lands Q1 2027.
- Claude Sheets: AI-generated multi-user spreadsheets. Describe what you need in plain English and Claude produces a full spreadsheet with formulas, pivot summaries, and conditional formatting. Subsequent edits are conversational ("add a 90-day rolling average column"). Exports to Excel and Google Sheets.
- Claude Data: Connects Claude to data warehouses (BigQuery, Snowflake, Redshift) and CRMs (Salesforce, HubSpot) to generate dashboards and data stories on request. Previously in limited beta as Claude Dashboards; now fully GA and renamed. The connector setup lives in Organization Settings → Integrations.
Admin controls to know
- Data: Default-on from November 14. To disable, go to Organization Settings → Cowork → Data and toggle off. Existing Dashboards beta users are automatically migrated.
- Whiteboards & Sheets: Off by default until November 20. Admins can enable now via Organization Settings → Cowork. After November 20 the defaults flip — opt-out before then if your compliance posture requires it.
- Data retention: All Cowork content is subject to the same data retention policies as Claude Projects. Organisations with zero data retention (ZDR) configurations should review whether Cowork use cases are compatible before enabling.
Who should enable Cowork today
Teams that already use Claude for document drafting will find the most immediate value in Sheets and Whiteboards — both surfaces reduce the copy-paste loop between Claude and external tools. Data is highest-value for teams with BI analysts who currently spend time translating stakeholder questions into SQL: direct-to-Claude connectors remove that translation layer. Pilot with a single team, review data-access scope, then roll out Organisation-wide.
Cowork
Whiteboards
Sheets
Claude Data
GA
Enterprise
Team
admin controls
🧭 Claude Code 2.1.325 — Security-Hardening Release: zstd Patch, Subagent Isolation, and Crash Recovery
Claude Code 2.1.325, released at ~20:40 UTC on November 13 and surfacing in most auto-update pipelines on the 14th, ships zero new features and four targeted security and reliability fixes. This is a mandatory-update-class release for anyone running autonomous Claude Code agents in production, particularly multi-agent pipelines where subagents operate with their own model configurations.
What changed
- zstd decoding vulnerability patched: A bug in Claude Code's compression pipeline allowed a malformed zstd-compressed tool response to cause a buffer overflow condition. The fix enforces strict size limits on decompressed payloads and rejects any response that exceeds them before parsing begins. Users who pass tool call results through custom MCP servers or proxy layers should verify their proxies are not injecting unusual compression headers.
- Subagent model handling hardened: Subagents with an explicitly configured
model field in their task definition were previously being coerced to the plan-mode model when Claude Code ran in plan mode. This caused model-identity confusion in multi-agent setups and could silently change cost and capability profiles mid-pipeline. The fix preserves the subagent's declared model throughout the session lifecycle.
- Checkpoint file isolation: Sandboxed subagent commands could previously read Claude Code's checkpoint file copies via path traversal within the sandbox mount. The fix restricts the sandbox mount to exclude the
.claude/checkpoints/ directory entirely for subagent-scoped sandboxes.
- Session-resume data loss fixed: Unclean exits (OOM kill, SIGKILL, sudden host shutdown) previously corrupted the session state file, causing the next resume to start from an empty context. Claude Code now writes session state to a temporary file and atomically renames it into place, so crashes during a write leave the previous state intact.
Update immediately if you run autonomous agents
The zstd vulnerability and checkpoint isolation fix are relevant to any deployment where Claude Code processes tool call responses from sources outside your direct control — for example, MCP servers operated by third parties, or agent pipelines where subagents fetch external content. Run claude update or let your package manager pull 2.1.325 before your next production agent run.
Claude Code
v2.1.325
security
zstd
subagent
sandbox
crash recovery
patch
🧭 November 13 Claude Login Outage — Resolved: 90-Minute Authentication Failure Traced to Config Rollback
On November 13, Claude.ai and Claude Code experienced a widespread login failure lasting approximately 90 minutes. Users attempting to sign in received "Authentication failed" errors on the web interface and "Unauthorized" (HTTP 401) responses from the CLI. Anthropic's status page confirmed the incident at 15:02 UTC and marked it resolved at 16:08 UTC. By November 14, all services are confirmed fully recovered and no residual impact has been reported.
What happened and what was affected
- Root cause: A backend infrastructure change earlier on November 13 introduced a misconfiguration in the authentication token-refresh endpoint. New logins and token-refresh attempts failed; already-active sessions were unaffected because they did not need to re-authenticate mid-session.
- Fix: Anthropic engineers identified the misconfiguration and issued a configuration rollback. No code changes were needed — the fix was a revert to the previous infrastructure state.
- Scope: Claude.ai web, iOS, Android, and the Claude Code CLI were all affected during the window. Enterprise SSO customers were routed through a separate authentication path and were largely unaffected, though some SSO providers that call back through the main token endpoint experienced partial disruption.
- Data exposure: No user data was exposed and no credentials were compromised. The outage was an availability failure, not a security breach.
CLI still showing authentication errors?
Some Claude Code users whose sessions were caught mid-refresh during the outage window may still see "Claude Code needs to be re-authenticated" errors even after the incident was resolved. Run claude auth login to establish a fresh session. If the error persists, run claude auth logout && claude auth login to clear any cached partial-token state.
outage
authentication
incident
Claude Code
claude.ai
status
resolved