X Ships an Official MCP Server: What XMCP Means for Your AI Coding Workflow

Until June 30, connecting an AI agent to X required you to build and host your own bridge: register a developer app, wire OAuth, write a FastAPI wrapper, keep your token refreshed, and hope X's API docs hadn't drifted. X just made all of that someone else's problem. XMCP is X's official hosted Model Context Protocol server — one endpoint at api.x.com/mcp, 200+ tools auto-generated from the X API OpenAPI spec, and your AI agent talking to a real-time social graph in under a minute.

X joins GitHub, Slack, Notion, Stripe, and Salesforce in shipping an official MCP server. The pattern is becoming the new baseline for how platforms distribute API access to AI agents.

What XMCP actually is

XMCP is two things shipped together: a hosted endpoint run by X at https://api.x.com/mcp, and an open-source self-host option at github.com/xdevplatform/xmcp. Both expose the same tool surface.

The tool surface is generated at startup from the X API OpenAPI specification — every operation in the spec becomes an MCP tool. That means post search, user lookup, trend retrieval, bookmark management, and more are all available as first-class tools your agent can call with structured inputs and get typed outputs from.

The transport is Streamable HTTP (MCP protocol version 2025-06-18), which means it works natively with any MCP client that supports the current spec: Claude Code, Cursor, Windsurf, Codex CLI, and anything built on the TypeScript or Python MCP SDKs.

Connect it in 60 seconds

The hosted endpoint authenticates with a Bearer token from your X developer account. Get one at developer.x.com under Projects & Apps → Keys and tokens → Bearer token.

# Add the X MCP server to your Claude Code project
claude mcp add x-api \
  --transport http \
  https://api.x.com/mcp \
  --header "Authorization: Bearer YOUR_BEARER_TOKEN"

For Cursor, Windsurf, or any IDE that reads an MCP config file, add a stanza to your .cursor/mcp.json or equivalent:

{
  "mcpServers": {
    "x-api": {
      "transport": "http",
      "url": "https://api.x.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_BEARER_TOKEN"
      }
    }
  }
}

Once connected, confirm the server is live by asking your agent to search a topic:

You: Search X for posts about the MCP spec release candidate from the last 7 days.

Agent: [calls search_posts tool via XMCP, returns structured results]

What you can query (and what you can't)

The 200+ tools cover most of the X API v2 surface. The categories that matter most for developer workflows:

• Post search: Full-archive search with filters by date, author, language, engagement threshold, and media type.
• User lookup: Profile data, follower counts, pinned posts — by username or user ID.
• Trend retrieval: Real-time trending topics by location (WOEID), returned as ranked lists your agent can reason about.
• Bookmark management: Read bookmarks from your authenticated account — useful for knowledge-base pipelines.
• Article drafting: Draft and publish long-form X Articles (write access requires OAuth 2.0 user context, not just a Bearer token).

X API cost reality check

There is no charge for the MCP layer itself — XMCP is free infrastructure on top of the X API. But every tool call your agent makes runs against your X API plan quota and pay-per-use credits.

For read-heavy workflows — monitoring mentions, trend dashboards, competitive analysis — the cost is negligible at most API plan tiers. Write-heavy agentic workflows (auto-drafting posts, bulk publishing) will accumulate credits faster. X's API dashboard shows per-endpoint usage in real time.

Real developer use cases

The combination of a structured MCP tool surface and X's real-time data opens up patterns that were annoying to build before.

Release monitoring

Ask your Claude Code agent to watch for posts mentioning a library or tool name, filter by engagement, and surface anything that looks like a bug report or breaking-change discussion. One prompt, runs in the background, surfaces findings into a Markdown file.

Search X for posts mentioning "@skills-hub-ai" OR "skills-hub.ai" from the last 48 hours.
Filter to posts with >10 likes or replies. Group by sentiment (positive / negative / question).
Write a summary to monitoring/x-mentions-$(date +%Y-%m-%d).md.

Trend-driven content briefs

Pull trending topics in your niche, cross-reference against your existing content, and generate a list of gap topics your next blog posts should address.

Competitive intelligence snapshots

Search for posts about a competitor, extract the top complaints and praise points, and output a structured report. This is the kind of market research that used to take a human analyst half a day.

Knowledge base ingestion

Fetch bookmarks from an authenticated account, parse each linked URL, and ingest the content into a local vector store. Keeps your agent up-to-date with what your team is reading without manual curation.

Why every platform will follow

X's decision to ship a hosted MCP server is part of a broader pattern. In 2026, the question for any platform with a developer API is no longer "should we build an MCP server?" — it's "how fast can we ship one before a third-party does it for us?"

The reason is distribution. When GitHub shipped its MCP server, it became trivially easy for every AI coding agent to read issues, check PR status, and trigger actions — without the developer writing a line of integration code. The platform that ships an official MCP server first gets the best agentic UX, attracts agent developers, and locks in usage before community forks establish the default.

For X, the calculus is clear: its most defensible asset is real-time data. XMCP turns that data into a first-class tool for AI agents at a moment when agentic workflows are the fastest-growing API usage category. The social media platforms that don't follow will find third-party MCP servers — with worse auth, stale schemas, and no SLA — becoming the de facto standard instead.

The skills-hub.ai x-mcp skill

The x-mcp skill on skills-hub.ai wraps the XMCP connection, authentication verification, and common query patterns into a single installable SKILL.md. It guides your agent through setup, confirms the connection, and provides phased instructions for monitoring, trend analysis, and content workflows.

# Install the x-mcp skill
npx @skills-hub-ai/cli install x-mcp

# Then ask your agent to use it:
# "Use the x-mcp skill to set up XMCP and monitor mentions of my project."

The skill is in the integration category alongside figma-mcp, mcp-enterprise-auth, and 20+ other official MCP server integrations.

Related reading: the MCP 2026-07-28 spec migration guide, the full MCP server catalog, and how claude mcp login makes MCP auth frictionless.