Claude Add-ins Now Share Context Across Apps and Support LLM Gateways

What Happened

On March 23, 2026, Anthropic shipped two meaningful changes to Claude's add-ins architecture. First, Claude add-ins now maintain full conversation context across multiple applications — meaning when you move from a research tool to a drafting tool to a project management add-in, Claude carries the full history of what was discussed instead of starting from scratch. Second, Anthropic added LLM gateway support — the ability for enterprises to access Claude through Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry, rather than exclusively through Anthropic's direct API. Both features are live now across Pro, Max, Team, and Enterprise plans.

What's New

• Cross-context sharing: Add-ins now share the full conversation history across applications. A planning conversation in one add-in is visible to another add-in in the same session.
• Skills within add-ins: Add-ins can now expose discrete, callable capabilities (skills) that other parts of a Claude workflow can invoke — enabling more modular tool design.
• LLM gateway support: Claude is now accessible through Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry — the three major cloud AI marketplaces.
• No plan changes: Cross-context sharing rolls out to all paid Claude tiers — Pro, Max, Team, and Enterprise. No pricing changes announced.

Why It Matters for Your Business

The context-switching problem is real and expensive. If you have used Claude across multiple tools in a single project — say, researching in a browsing add-in, drafting in a writing tool, and then creating tasks in a project management integration — you have probably spent time re-explaining context that Claude should already know. That friction is what this update removes.

Before this update, developers were solving the problem manually. Third-party tools like Mem0 were being used to sync context across ChatGPT, Claude, Perplexity, and other assistants via browser extensions and custom scripts. The fact that a workaround ecosystem existed confirms how badly this was needed. Native cross-context sharing eliminates the need for those workarounds in Claude-centric workflows.

Here is what each audience gains right now:

• Agencies: Client research done in one add-in now feeds directly into brief drafting and task creation in others. No more paste-and-explain.
• SMB teams: A single Claude conversation can now drive actions across tools like Asana, Slack, and Figma without manually bridging the context gap between each.
• Builders and indie developers: You can now build add-ins that assume the full conversation history is available — no need to re-prompt users or pass context manually between tools.
• Enterprise teams on AWS, GCP, or Azure: You can now route Claude through your existing cloud infrastructure instead of managing a separate Anthropic API relationship. If your organization has standardized on Bedrock or Vertex AI, Claude access becomes one less exception to manage.

The LLM Gateway Play: Why This Is a Big Deal for Enterprise

An LLM gateway is a managed cloud service that gives enterprise teams access to third-party AI models through their existing cloud provider — rather than requiring a direct integration with each model vendor. Think of it like how AWS Bedrock already lets you access models from Cohere, Meta, and Mistral without dealing with each company's API separately. Claude is now in that category on all three major platforms.

For organizations that have already negotiated enterprise agreements with AWS, Google Cloud, or Microsoft, accessing Claude through those agreements may reduce effective per-token costs compared to a separate Anthropic API contract — though Anthropic has not published a direct pricing comparison. It also consolidates billing, vendor management, and compliance review into infrastructure teams already maintain.

The MCP Strategy Behind the Feature

Cross-context sharing and LLM gateway support are both built on Anthropic's open-source Model Context Protocol (MCP). MCP is an open standard that lets tools expose both their functionality and their interactive UI so that any AI system supporting the protocol — not just Claude — can host them. Anthropic has publicly framed this as: "A tool built using MCP can expose both its functionality and its interactive UI through an open standard. Any AI system that supports MCP — not just Claude — can then host that same tool without requiring a custom, platform-specific integration."

The strategic bet here is significant: by making MCP open and not Claude-exclusive, Anthropic is positioning it as a potential industry standard rather than a proprietary moat. If MCP becomes the default protocol for AI tool interoperability — the way REST became the default for web APIs — Anthropic's early position gives it a durable advantage in how the ecosystem is shaped, even if competitors adopt the same standard.

The Bigger Picture: Where Claude Is Headed

This update is the latest step in Anthropic's consistent push to make Claude a primary workspace rather than a supplementary tool. Recent milestones — MCP Apps (interactive tool embedding), expanded tool connectivity, and now cross-context sharing and gateway support — form a coherent pattern. Anthropic is systematically removing the friction points that push users back to manual workarounds.

The competitive read: OpenAI's ChatGPT has had persistent context and tool use for longer, but operates primarily within OpenAI's ecosystem. Google Gemini integrates well with GCP but is less useful for multi-cloud organizations. Microsoft Copilot is tightly bound to Azure and M365. Claude's combination of open MCP, multi-cloud gateway access, and cross-context sharing is a direct response to all three — and positions it as the most cloud-agnostic option for enterprise teams that don't want to bet on a single vendor's stack.

What to Do Right Now

• If you are on a paid Claude plan (Pro or above): The cross-context feature is already available. Audit any workflows where you are currently copy-pasting context between add-ins and test whether native sharing eliminates the step.
• If you are building Claude add-ins: Review the MCP documentation to understand how skills and cross-context sharing work architecturally. Add-ins that previously had to request context re-entry from users can now be redesigned to assume availability of full session history.
• If you are an enterprise IT or procurement team: Evaluate whether accessing Claude via your existing Bedrock, Vertex AI, or Foundry agreement simplifies vendor management or reduces cost compared to a direct Anthropic API contract.
• If you are using Mem0 or a similar context-sync workaround: Test whether native cross-context sharing covers your use case before maintaining the additional dependency.