X launches an MCP server — platform data now available to AI agents

X on June 30, 2026, launched a hosted Model Context Protocol (MCP) server that lets AI tools — including Claude, Cursor, and Grok Build — communicate with the platform's API using a user's own account permissions. Previously, developers who wanted an AI assistant to access X data had to build, host, and maintain their own MCP server.

Key takeaways

X now hosts its own MCP server — developers no longer need to build custom integrations with the X API
Authentication runs through the user's X account — AI apps operate within the permissions the user already has
Access includes: X search, post reading, user profile lookup, conversation and trend analysis
No access to the Write API — AI tools cannot post to X through this server
X joins GitHub, Slack, Notion, Stripe, and Salesforce in offering official MCP servers

MCP as the standard for connecting AI agents to external data

Model Context Protocol is an open standard developed by Anthropic that defines a common way for AI models to communicate with external tools and services. Rather than each provider building a custom integration for every AI assistant, MCP standardizes that layer: one protocol, one API shape, many AI clients.

Until now, X had an API accessible to developers, but anyone who wanted to connect an AI assistant like Claude Code or Cursor to X had to stand up their own MCP server, configure OAuth authentication to the X API, and maintain the full infrastructure stack themselves. X is now taking that work off developers' hands.

What you can and cannot do with X MCP

The hosted MCP server gives access to the same capabilities that have long existed in the X API: searching posts, reading tweets, looking up user profiles, and analyzing conversations and trends. It does not add new API endpoints — it lowers the integration barrier.

X confirmed to TechCrunch that the MCP server is not compatible with X's Write API endpoints. AI tools cannot post or publish content to X through this server — neither interactively nor autonomously.

That is a deliberate guardrail in the context of spam concerns. X also updated its API v2 earlier in 2026 specifically to address AI-generated replies in conversations. The cost of publishing links through the API rose to $0.20 per post.

Official MCP servers as a platform strategy

X is not the first service to launch its own official MCP server. GitHub, Slack, Notion, Stripe, and Salesforce all did so before. Each platform offering a native MCP integration signals that MCP is becoming the de facto standard for exposing external data to AI assistants.

For X, the MCP server carries an additional message: it positions Elon Musk's platform not as a social network but as a real-time data source that AI agents can query and analyze.

Why this matters

MCP is becoming the connective layer between AI agent ecosystems and external data. As more major platforms launch official servers, the barrier for developers building agentic applications drops and data access becomes more unified.

X brings assets most platforms lack: real-time data, public conversations, and profiles of people who matter in public life. For AI agents that need current social or media context, that is a useful source — as long as data quality holds up for developers.

The read-only constraint (no Write API) is a considered decision: X lowers friction for analytical tools while protecting the platform from autonomous content generation by agents acting on its behalf.

What's next

MCP server documentation is live at docs.x.com — developers can integrate it with Claude, Cursor, and other MCP clients immediately
Write API access remains outside the scope of MCP — further expansion is possible but X has not announced any plans