X launches a hosted MCP server so AI assistants like Claude and Cursor can plug directly into its platform

X has launched a hosted Model Context Protocol (MCP) server, letting AI assistants and coding tools connect directly to X's platform and API without developers having to build or maintain their own integration layer.

What's new

Per TechCrunch's report on the launch, MCP "is an open standard that defines a common way for AI models to connect to external tools and services," and X's new offering hosts that connection itself: "The hosted MCP doesn't add new capabilities on that front; it just makes them easier to expose to AI applications." Once connected, tools such as Claude Desktop, Cursor, and Grok Build gain programmatic access to X features including full-archive search, trending topics, and bookmarking, operating under the connecting user's own account permissions.

Access is read-only for posting purposes. TechCrunch reported that "X clarified to TechCrunch that the MCP tool is not compatible with X's Write API endpoints, so it's not possible to use it to post autonomously (or at all) on X" — meaning the integration is scoped to retrieval and analysis rather than enabling autonomous posting or spam generation.

Context

MCP, originally developed by Anthropic, has rapidly become a common integration standard across the AI tooling ecosystem over the past year, with GitHub, Slack, Notion, Stripe, and Salesforce among the companies that have shipped official MCP servers for their platforms. X's move brings its real-time social data — full-archive search, trends, and bookmarks — into that same standardized integration layer, rather than requiring each AI tool vendor to build and maintain a bespoke X integration.

Why it matters

Hosting the MCP server itself, rather than leaving third parties to reverse-engineer or maintain unofficial connectors, gives X more control over how AI agents access its data while lowering the integration cost for tool builders. It's also a sign of how central live social and web data has become to agentic AI workflows: coding assistants and general-purpose agents increasingly need to search, monitor, and reason over real-time information, not just static documentation, and platforms that ship first-party MCP access position themselves as a default data source for that emerging agent economy. The read-only restriction is a deliberate guardrail against the more obvious abuse case — autonomous agents mass-posting or manipulating platform activity — while still opening the door to legitimate research, monitoring, and analysis use cases built on top of X's data.