Anthropic gives Claude hands to work across your work apps

If you have tried using Anthropic's Model Context Protocol over the past few months, you know the drill: great tech, but trapped inside local config files on Claude Desktop. Switch over to the browser, and Claude was suddenly blind to your workspace again.

That limitation is finally gone. Anthropic just introduced Integrations, bringing remote MCP server support directly to both the web and desktop versions of Claude. It connects your everyday work tools straight into the model without requiring you to run a terminal in the background.

Remote MCP goes live with 10 heavy hitters

When Anthropic introduced MCP last November, it was an open standard meant to connect AI apps to data, but it only worked through local servers on Claude Desktop. Integrations changes that completely by allowing Claude to talk to remote MCP servers hosted anywhere.

At launch, Anthropic rolled out integrations for 10 widely used business platforms: Atlassian's Jira and Confluence, Zapier, Cloudflare, Intercom, Asana, Square, Sentry, PayPal, Linear, and Plaid. Upcoming integrations from Stripe, GitLab, and Box are already lined up.

This turns Claude into an active teammate rather than just a chat window. Plug in Zapier, and Claude can pull sales metrics from HubSpot or assemble meeting prep straight from your calendar. Connect Jira and Confluence, and it can draft specs while spinning up related project tickets simultaneously. Intercom's Fin AI agent even acts as an MCP client, automatically logging software bugs in Linear whenever customers report issues.

For engineering teams wanting custom setups, Anthropic says developers can build a working integration in about 30 minutes, especially when using Cloudflare's pre-packaged OAuth and deployment tooling.

A 45-minute deep research engine

Alongside tool integration, Anthropic expanded its Research capabilities. Claude can now perform complex investigations that pull across the open web, your Google Workspace, and any connected Integrations at the same time.

Instead of rushing to deliver an instant answer, flipping on the Research toggle instructs Claude to dissect your prompt into granular research threads. While typical reports finish in 5 to 15 minutes, Claude can dig across hundreds of sources for up to 45 minutes on dense questions before returning a formatted brief with direct source citations.

Both Integrations and advanced Research are available for Pro, Max, Team, and Enterprise subscribers, while basic web search is now accessible to all Claude users worldwide.

Why it matters

Anthropic is quietly solving the single biggest annoyance with generative AI: the endless copy-pasting between browser tabs. By moving MCP from a geeky local prototype to managed cloud integrations, Claude transforms from a text generator into a command center for your entire SaaS stack.

If your daily grind involves jumping between Linear tickets, customer support logs, and internal Confluence pages, this is well worth testing immediately. Letting an AI synthesize unstructured feedback and turn it into actionable backlog items in one prompt saves genuine hours.

We think Anthropic made a brilliant strategic bet by pushing MCP as an open standard rather than a proprietary walled garden. Partnering with Cloudflare for authentication and Zapier for instant access to thousands of niche apps gives them an immediate distribution moat that competitors will struggle to match.

That said, enterprise security teams are bound to lose some sleep. Handing an AI direct read and write hooks into corporate issue trackers, payment processors, and customer chats raises the stakes dramatically. If Claude misinterprets an action or hallucinates an update inside a mission-critical tool, the blast radius is much bigger than a sloppy email draft.