Until now, a typical generative AI workflow has been a one-off task: ask the AI to produce a table or file, download it to your PC, and edit it in another app. On November 10, 2026, Anthropic announced a set of new features that significantly changes that approach. It introduced "Claude Sheets," a spreadsheet tool that teams can edit together directly inside Claude, and "Claude Whiteboards," a space for organizing ideas. It also added a mechanism that displays the interfaces of external services inside Claude so users can operate them in place.
The announcement suggests that Anthropic wants to develop Claude from an AI that answers questions into an integrated work environment for handling business data and using a variety of external services. However, many of the announced features are in beta, and quite a few offer fewer capabilities than existing business software. Some features require activation by an administrator, and some are unavailable in regulated environments. Companies considering adoption need to understand exactly which features are available and what limits apply today.
A Mix of Beta and Generally Available Features
Not all of the new features Anthropic announced are at the same stage. Some have been released as betas, some have moved to general availability, and some have been announced for future release. Available plans and the initial settings administrators must configure also differ, so organizations considering company-wide rollout need to check the conditions for each.
データを表で見る
| 上限・登録規模 (件 / セル) | |
|---|---|
| Marketplace提供数 | 9,000 |
| Sheets上限セル数 | 100,000 |
Claude Whiteboards and Claude Sheets, the centerpieces of this announcement, are both offered as betas. Meanwhile, the data visualization feature previously known as "Dashboards" has been renamed "Claude Data" and has reached general availability (GA). Even so, advanced Claude Data features such as certified queries and live sharing remain in beta. Anthropic also previewed "Claude Meetings," which joins voice calls to help produce meeting notes and update documents, but has not disclosed a launch date.
According to Anthropic, the three products (Whiteboards, Sheets and Data) are available on all paid plans. The published materials do not mention optional add-ons that cost extra or regional availability restrictions. The products support web, desktop and mobile, though specific operations and features vary by environment. To use Claude Data, users must also connect data sources such as data warehouses and customer relationship management (CRM) systems through connectors.
Default settings also differ across organizational plans. On the Team plan, all three products are enabled from the start. On the Enterprise plan, however, Whiteboards and Sheets are disabled by default and administrators must enable them manually. Claude Data on the Enterprise plan, meanwhile, is scheduled to be enabled automatically on November 19, 2026. Administrators need to take these differences into account when adjusting rollout timing and user permissions.
| 製品・機能 | 提供状況 | 対象プラン・利用環境 | 主な条件・制約 |
|---|---|---|---|
| Claude Sheets | ベータ版 | 全有料プラン。Web・デスクトップで編集、モバイルでは閲覧中心 | 1シート10万セル。ピボットテーブルやマクロ非対応。HIPAA/CMEK/ZDR対象外 |
| Claude Whiteboards | ベータ版 | 全有料プラン。Web・デスクトップ・モバイル | チームでの共同編集、アイデアの自動整理、外部ツールへの書き出しに対応 |
| Claude Data | 一般提供（一部機能はベータ） | 全有料プラン。データコネクターが必要 | データウェアハウスやCRMとの接続が必須。認定済みクエリとライブ共有はベータ版 |
| 対話型コネクター | 順次提供中 | 全プラン。Web・デスクトップ | 新しいClaudeのUIを有効化する必要がある。FigmaやShopifyなどは開発中 |
| Claude Meetings | 提供予定 | 未発表 | 通話への参加や文書更新を予定。提供時期は未公表 |
In short, this announcement is not the launch of a single, finished business suite but a staged rollout of multiple features. In Enterprise environments in particular, activation timing and conditions of use differ by feature, so it is important for administrators to check settings and internal policies before deployment.
Claude Sheets' 100,000-Cell Limit and How It Differs from Existing Spreadsheet Software
The newly introduced Claude Sheets is not merely a feature in which AI generates tables. It supports multiple sheet tabs, formulas that span tabs, formatting, conditional formatting, dropdown lists, checkboxes and charts. Multiple people can edit the same sheet together, and cells changed by Claude are marked with a border.
That said, it is not a drop-in replacement for Microsoft Excel or Google Sheets. Anthropic itself positions Claude Sheets as a tool suited to lightweight shared spreadsheets. For more advanced spreadsheet needs, it directs users to "Claude for Excel" or "Claude for Google Workspace," or to export the file and edit it in existing software.
The major practical constraint is the limit of 100,000 cells per sheet. With 10 columns of data, for example, that limit is reached at 10,000 rows. For uses involving large volumes of transaction history or inventory data, the limit could be reached fairly easily. In addition, the following features, widely used in existing spreadsheet software, are not supported in the initial release:
- Pivot tables
- Protection of cell ranges
- Find and replace
- Sorting by multiple columns
- Filter views
- Multi-select dropdowns
- Export to PDF
Macros, VBA, Power Query, DAX and print layout settings are also unsupported. Data can be imported in CSV, Google Sheets or Excel formats, and created sheets can be exported as .xlsx or CSV. Because Sheets usage counts against the normal usage allowance of the subscribed plan, usage volume also deserves attention when processing large amounts of data continuously.
Another important point concerns limits on integration with external data. At present, Claude Sheets cannot automatically pull the latest data from connected external apps and continuously update sheet contents. Nor is there a feature that lets documents, slides or dashboards created inside Claude directly reference Claude Sheets as a data source. Anthropic describes the ability to carry work context across tools as a characteristic of the new Claude, but for Sheets at least, it is closer to a tool for collaborating on imported data than a business foundation that stays constantly in sync with external data.
There are also restrictions on the usage environment. Sheets can be edited on the web and desktop versions, but the mobile version is basically limited to viewing. In addition, during the beta period Claude Sheets does not support the U.S. health information privacy law HIPAA, customer-managed encryption keys (CMEK) or zero data retention (ZDR). As a result, it is difficult to adopt the current version of Sheets as is for medical information or highly sensitive business data that must meet these requirements.
Claude Whiteboards, meanwhile, focuses on helping teams brainstorm and organize discussions. Multiple people can work in real time, and Claude can group related ideas and extract next actions. Created content can be converted into documents or slides, and exported to external tools such as Lucidchart, Miro and Zoho's Vani so work can continue there.
Interactive Connectors and MCP Apps Bring External App Operation into Claude
The announcement also significantly expanded how Claude connects to external services. At the center is "Interactive connectors." Whereas conventional connectors mainly retrieved data from external services and executed APIs, the new mechanism can display the operating interface of an external app inside Claude itself.
The technical foundation is the Model Context Protocol (MCP), which Anthropic released as open source on November 25, 2024. MCP is a protocol that lets AI models use external data sources and tools through a common mechanism. Interactive connectors extend this connection mechanism to the user interface, allowing external services to be operated directly within Claude's work screen.
Users can use external services beside the Claude chat or in full screen without switching to another app. Authentication carries over the user's own sign-in state, and when Claude performs an operation that makes changes, an approval process is in place just as with conventional connectors. While AI can now operate external services, the design aims to prevent unintended changes through user confirmation.
Interactive connectors are being rolled out gradually on web and desktop for all plans. However, using them requires that the new Claude UI be enabled, and depending on organizational settings, administrator approval may also be required. Even once the feature has started rolling out, not every user can necessarily use it immediately.
Supported services include Asana, Canva, Carta, Dropbox, monday.com, Smartsheet, Vercel and Webflow. In the same announcement, Anthropic also showed examples using Adobe, Atlassian, Box, HubSpot, Notion and others. Integrations for Figma, Harvey and Shopify, on the other hand, are explicitly stated to be under development, so they should be distinguished from services available at the time of the announcement.
For developers, a public preview of "MCP Apps," an extension specification of MCP, has also begun. It allows third-party developers to build their services' features and embedded interfaces as apps that run inside Claude. A distinguishing point is that Claude can serve as the starting point for operations, rather than merely calling external services.
A precursor to this approach is "Salesforce in Claude," offered separately as a beta. Using it, however, requires a paid Claude plan, Salesforce-side approval and a contract for the latest "Sales Cloud Enterprise edition." It combines Salesforce and Slack connectors and provides 37 skills for sales work.
Write operations such as updating Salesforce data require user approval by default. An always-allow setting can also be chosen, so organizations need to establish appropriate operating rules. In addition, the data and operations accessible through Claude are limited to each user's existing access permissions in Salesforce.
Anthropic's marketplace lists more than 9,000 connectors and plugins, including community-developed ones. However, not all have undergone full review by Anthropic. Nor can all 9,000 be used as interactive connectors with embedded interfaces, so the number of listings should be distinguished from the range of features actually available.
Enterprise Authentication Management and Isolated Environments Come with Conditions
The more external services can be connected to Claude, the more important access permissions and security management become for companies. Anthropic also explained that it has strengthened Enterprise management features and execution-environment isolation. However, because security and conditions of use differ by authentication method, it is necessary to understand how each feature works.
Enterprise authentication management is available on the Team and Enterprise plans. Authentication via an identity provider (IdP) is generally available, while Claude-issued authentication remains in beta. Using IdP-managed authentication requires single sign-on (SSO) integration with a supported IdP, and at the time of the announcement the only formally supported provider is Okta.
Particular caution is needed with the beta Claude-issued authentication. This method uses only an email address to match users. Claude confirms that the user is a member of the organization, but does not check SSO, multi-factor authentication (MFA) or the sign-in conditions set by the connected service itself. Depending on how this is combined with access controls on the external service side, access may therefore remain possible through routes a company did not anticipate.
Regarding connector use, administrators can enable features according to user roles, limit the scope of access permissions and revoke permissions already granted. However, connectors and external IdPs are third-party services with their own terms of use and management structures. Even if use can be controlled on the Claude side, Anthropic does not assume all security and legal responsibility related to external services.
For the execution environment, the new unified Enterprise environment runs each AI agent session on an independent virtual machine (VM). Communication to external networks is blocked by default, and authentication credentials for external tools are managed in isolation outside the sandbox. This separates the processing the agent performs from the credentials, and is designed to reduce the risk of unintended access or information leakage.
According to Anthropic, Enterprise plan data is encrypted both in transit and at rest and is not used for AI model training by default. There are exceptions, however, to migration to this new unified Enterprise environment. Organizations that apply zero data retention (ZDR) in Claude Code, those using HIPAA-compliant configurations and those that have adopted customer-managed encryption keys (CMEK) are not subject to migration to the new unified environment and will continue to use the conventional configuration.
On the auditing side, audit logs and a Compliance API are also provided for Enterprise. Anthropic says it can integrate with more than 100 security and compliance tools, including Splunk and CrowdStrike. However, the ability to obtain audit logs and integrate with external tools is a separate matter from guaranteeing conformity with each company's regulatory requirements.
How Far Can Practical Adoption Go?
As Claude evolves into a work environment that handles business more broadly, where should companies begin introducing the new features? For now, it is more realistic to start with uses where feature limitations have a small impact on work, rather than replacing existing core systems. Examples include team brainstorming, early-stage document drafting, short-term progress trackers and visualization of approved data sources. Such tasks are easier to complete within a single work environment and depend relatively little on continuous synchronization with external data or complex permission management.
Anthropic also described results from its own use. In a Slack-integrated tool built by its legal team, the time required for marketing-related contract review was reportedly cut from the previous three days to 24 hours.
データを表で見る
| レビュー時間 (時間) | |
|---|---|
| 導入前 | 72 |
| 導入後（社内報告） | 24 |
However, this is an internal case reported by Anthropic itself, not an independent evaluation by a third party or a comparative experiment under uniform conditions. It is notable that legal and finance staff were able to build business tools without relying on specialized development experience, but other companies will not necessarily achieve the same results. If the structure of the data handled, permission management and audit procedures differ, the development work required and the benefits of adoption will also change.
Caution is also needed when adopting plugins. Claude plugins combine skills, connectors, subagents and other elements, but hooks and subagents run not in ordinary chat but in the "Cowork" and "Claude Code" environments. Anthropic warns that plugins may include MCP servers that run with the same privileges as local programs. In Enterprise environments, it is therefore worth considering restricting plugins from untrusted providers or disabling the execution of local MCP servers altogether.
When companies proceed with pilot deployments, it is important to verify not just how new the features are but whether they can be used safely and reliably in actual work. In particular, the following five points should be verified before adoption:
- Calculation accuracy and file compatibility: Do formulas calculate correctly in Claude Sheets, and are formatting and calculation results preserved when exported to Excel or similar software?
- Approval flow for operations: Is appropriate approval required for writes and data updates to external services, preventing mistaken operations and unintended changes?
- Principle of least privilege: Are connectors granted any unnecessary access permissions? Can the permissions of departed or reassigned employees be revoked properly?
- Requirements for integration with external data: Can the work be done with temporarily imported data alone, or does it require automatic synchronization with the latest data?
- Legal and compliance requirements: Does the work involve data that requires HIPAA compliance, CMEK, ZDR or the like? Do the features in use meet the company's security standards?
The announcement made clear that Anthropic intends to develop Claude from a simple AI chat service into an integrated work environment for day-to-day business. The direction of consolidating spreadsheets, collaborative editing and external app operation into a single screen could influence how conventional business software is used. At the same time, challenges remain before companies can adopt it fully: limited spreadsheet functionality, gaps in external data integration, differences between authentication methods and exemptions for regulatory compliance. Going forward, the focus will be on how far these limits are resolved as beta features reach general availability, and whether Claude can connect safely with existing business systems.