Lead
On Monday, Anthropic rolled out a new capability that lets Claude users open interactive third-party apps inside the chatbot interface, enabling direct actions in workplace services. The initial catalog centers on productivity tools such as Slack, Canva, Figma, Box and Clay, with a Salesforce connector described as arriving soon. The integrations permit an authenticated service instance to be accessed by Claude so users can send messages, generate visuals or retrieve cloud files from inside a chat session. The feature is available to Pro, Max, Team and Enterprise subscribers and must be enabled at claude.ai/directory; free users are excluded.
Key Takeaways
- Anthropic launched in-chat interactive apps for Claude on January 26, 2026, aimed mainly at workplace workflows and productivity.
- The first roster includes Slack, Canva, Figma, Box and Clay; a Salesforce integration is expected soon, per Anthropic’s announcement.
- Apps operate against a logged-in instance of each service so Claude can act with user-authorized access to send messages, edit graphics or fetch files.
- The capability is offered to paid tiers—Pro, Max, Team and Enterprise—and is not available to free accounts; users enable tools via claude.ai/directory.
- These integrations are built on the Model Context Protocol (MCP), an open standard Anthropic introduced in 2024 that added app support in November.
- The apps are intended to extend Claude Cowork, Anthropic’s multistage agent tool launched the prior week, though app access in Cowork is not available at initial release.
Background
Anthropic has positioned Claude toward enterprise and workplace customers since expanding its product lines, investing in features that tie model capabilities to company systems. The company introduced MCP in 2024 to make it easier for models and external tools to exchange structured context; MCP added explicit app support in November. OpenAI and other vendors have pursued similar models of interactive third-party tools—OpenAI’s Apps system launched in October—so this rollout follows an industry trend toward embedding task-specific utilities inside chat models.
Anthropic’s strategy emphasizes controlled, authenticated integrations for business workflows: by requiring logged-in service instances, the company aims to balance convenience with access controls. The first set of partners focuses on widely used workplace services—messaging, design, file storage and contact/data platforms—because those categories map cleanly to common tasks users ask Claude to perform. At the same time, Anthropic has been developing agentic features such as Claude Cowork that let users delegate multistage work, which increases the appeal of direct app access.
Main Event
On January 26, Anthropic announced that eligible Claude users can enable interactive apps inside the chat UI. When a user authorizes a service, Claude can interact with that service’s logged-in instance to carry out actions like sending a Slack message, editing a Figma design or pulling files from Box. Anthropic emphasized that the experience combines a visual interface—the native app screens—with Claude’s reasoning, allowing faster iterations than either can provide alone.
The launch list currently includes Slack, Canva, Figma, Box and Clay; Anthropic said a Salesforce connector is expected shortly. Eligible subscribers can activate available tools via claude.ai/directory, and Anthropic has indicated that more partners will be added over time. The company noted that while apps are immediately available within regular Claude sessions, they are not enabled inside Claude Cowork at this moment; Cowork will receive app access in a forthcoming update.
Anthropic framed the integrations as productivity accelerants but also reiterated safety advice for agentic systems. Its documentation for Cowork cautions users to limit permissions, avoid exposing sensitive documents and to consider creating dedicated folders or restricted access scopes for Claude to use. Anthropic advised close monitoring of agent actions and minimizing broad permissions to protect credentials and personal records.
Analysis & Implications
The integrations signal Anthropic’s intent to deepen Claude’s role inside enterprise workflows, not just as a conversational assistant but as an integrated executor of tasks across services. By enabling authenticated, in-context access to widely used workplace apps, Anthropic reduces context switching—users no longer need to copy text between tools or manually translate Claude outputs into platform-specific actions. For companies, that can translate to faster content iterations, quicker data retrieval and tighter alignment between AI outputs and operational systems.
Dependence on an open standard like MCP carries strategic consequences. MCP’s open design lowers friction for third-party developers and encourages interoperability between vendors’ app ecosystems, which benefits customers who want multi-model or multi-vendor workflows. At the same time, it raises the bar on governance: organizations must define clear permission boundaries and auditing for model-driven actions across connected services. The stewardship and adoption of MCP will influence whether integrations remain composable or become siloed within a vendor’s ecosystem.
From a competitive perspective, Anthropic’s move mirrors product decisions by other major AI platforms and highlights a race to become the central interface for knowledge work. Customers choosing a primary assistant will weigh the breadth of available apps, the quality of connectors, and the vendor’s track record on safety and enterprise controls. For Anthropic, execution matters: delivering reliable, secure connectors and timely Cowork integration will determine whether these features drive meaningful adoption among corporate teams.
Comparison & Data
| App | Primary Claude capability | Availability at launch |
|---|---|---|
| Slack | Send messages and interact with channels | Yes |
| Canva | Generate and iterate visual designs | Yes |
| Figma | Update and edit design files | Yes |
| Box | Access and retrieve cloud-stored files | Yes |
| Clay | CRM and contact data actions | Yes |
| Salesforce | CRM actions (expected) | Coming soon |
The table shows the first-wave focus on messaging, design and file-access platforms—categories that commonly appear in enterprise workflows. By contrast, earlier assistant app programs emphasized single-purpose utilities; Anthropic’s selection targets collaborative and persistent work artifacts (design files, message histories, document stores). The practical effect is that Claude can now materially participate in ongoing projects rather than only supply static outputs.
Reactions & Quotes
Anthropic framed the feature as a way to combine visual tools with model intelligence to speed workflows. The company highlighted the complementary nature of dedicated interfaces and Claude’s reasoning when announcing the feature.
“Analyzing data, designing content, and managing projects all work better with a dedicated visual interface,”
Anthropic (blog post)
Following that, Anthropic reiterated safety guidance for systems that can act autonomously across services, urging restraint in permission grants and recommending that users create limited-access working folders for agent use.
“Be cautious about granting access to sensitive information like financial documents, credentials, or personal records,”
Anthropic safety documentation
Independent analysts and enterprise security leads have noted that these kinds of integrations offer clear productivity benefits but increase the need for robust identity, access and audit controls. Organizations considering adoption will likely pilot with low-risk data sets and narrow scopes before wider rollouts.
Unconfirmed
- Exact timing for the Salesforce connector rollout is not specified by Anthropic beyond a general “soon” and lacks a firm date.
- Details on the depth of each app’s access model (read-only vs. write capabilities) vary by partner and were not fully enumerated in the announcement.
- Although Anthropic said Cowork app integration is coming, the timeline and permission model for agent-enabled app actions inside Cowork remain unannounced.
Bottom Line
Anthropic’s in-chat app integrations bring Claude closer to becoming a practical executor inside enterprise workflows by enabling authenticated, in-context actions across common workplace services. For paid subscribers, the new capability can reduce manual handoffs and speed iterative tasks in design, messaging and file retrieval—areas where context continuity matters most. However, the feature’s long-term value will depend on the breadth and quality of connectors, the safety model around permissions, and how quickly Cowork and other agentic tools safely adopt app access.
Enterprises should evaluate the feature by piloting with non-sensitive datasets, defining strict access scopes and ensuring audit trails for model-driven actions. If MCP gains broader adoption, organizations may benefit from reusable integrations across multiple assistants; if governance is weak, the same connections could increase exposure. Careful rollout, clear policies and vendor transparency will determine whether these integrations become a productivity advantage or a governance headache.