Twitter/X MCP servers compared: what each one can do

By Naod Tadele · October 28, 2026 · 6 min read

If you code with an AI tool, it's natural to want it to post about the work too. MCP (Model Context Protocol) is how tools like Claude Code, Codex, Cursor and VS Code talk to outside services, and there are a few MCP servers for X.

They are not the same kind of thing. Some hand your AI the raw X API. Some give it a couple of simple actions. Some sit between your AI and X and keep you in the loop. This guide covers what each one does, based on its own documentation, so you can pick the one that fits how you work.

The three kinds of X MCP servers

Before comparing names, sort them by shape:

1. Full X API servers. Your AI can call most of the X API: post, delete, like, follow, search, send DMs. Powerful, and it acts directly on your account.
2. Minimal posting servers. A small set of tools, usually "post" and "search". Easy to understand, still posts directly.
3. Draft-first servers. Your AI writes a draft, and a human approves it before anything reaches X.

Most of the choice comes down to one question: do you want your AI to be able to publish on its own?

xmcp

xmcp is published under the xdevplatform organization on GitHub. Its README describes it as a local MCP server that exposes the X API OpenAPI spec as tools, using FastMCP. Streaming and webhook endpoints are left out.

What it can do:

• Posting and engagement: tools like createPosts, deletePosts, likePost and repostPost.
• Search: recent and full-archive post search, user search and more.
• Accounts and graph: look up users, follow, mute and block.
• Everything else in the spec: lists, direct messages, Spaces, trends and media upload, among others.

What it needs:

• Python 3.9 or newer, started locally with `python server.py`.
• Your own X developer app: a consumer key, consumer secret and bearer token.
• An OAuth consent step in your browser when the server starts. The README says tokens are kept in memory only while the server runs.

By default it serves MCP over HTTP at a local address on port 8000. You can limit which tools load with an allowlist setting, and you should: an AI with deletePosts and followUser on your real account is a lot of trust.

Good fit if you are building on the X API, want your agent to read and search X, or want full control and are comfortable running and securing a local server.

twitter-mcp

twitter-mcp is a small open source server you run with npx. It exposes two tools:

• post_tweet: posts a new tweet.
• search_tweets: searches tweets.

It needs four credentials from a Twitter/X developer account (API key, API secret, access token and access token secret), passed as environment variables. The README walks through setup for Claude Desktop.

Good fit if you want the simplest possible "post this" and "find that" from your AI, and you're fine with it posting straight to your account when you ask.

Loudcode

Loudcode's MCP server is draft-first. It is a remote server, so there's nothing to run locally: you add one address and an API key to Claude Code, Codex, Cursor, Windsurf, VS Code (Copilot) or Cline. You don't need your own X developer app.

What it can do:

• draft_post: writes a post about the work you just did, in your voice, and saves it as a draft. It's not posted.
• list_drafts: shows what's waiting for review and what's queued.
• queue_post: queues a draft for your next free posting slot, or a time you give, only when you ask.
• post_metrics: views, likes and replies for your recent posts.
• list_ideas: post ideas from your product, numbers and repo for quiet weeks.
• weekly_checkin: answers the weekly check-in from what your AI knows and plans a week of drafts.

What it can't do: read your timeline, search X, reply, like, follow or send DMs. It also only posts to X today.

Good fit if you want to say "post about what we just shipped" at the end of a session and review the result before it goes anywhere.

Side by side

• Where it runs: xmcp runs locally (Python). twitter-mcp runs locally (Node via npx). Loudcode is a hosted HTTP server.
• Credentials: xmcp and twitter-mcp need your own X developer app keys. Loudcode needs a Loudcode API key, and you connect X once inside Loudcode.
• Scope: xmcp covers most of the X API. twitter-mcp covers posting and search. Loudcode covers drafting, queueing and metrics.
• Who publishes: with xmcp and twitter-mcp, your AI can post directly. With Loudcode, you approve every post.

How to choose

1. You want your agent to read X (search, monitor mentions, research): you need a full API server like xmcp.
2. You want one-line posting and nothing else: a minimal server like twitter-mcp is the smallest setup.
3. You want to post about your work without your AI speaking for you unchecked: pick a draft-first server.

You can also run more than one. A sensible split: a full API server for research, and a draft-first one for anything that goes out under your name.

Where Loudcode fits

Loudcode is built for the third case. Your AI already knows what you shipped in the session, so it hands Loudcode a summary and Loudcode writes the post in your voice, using your Voice & tone settings. The draft lands in your queue, where you edit it, schedule it or throw it away. Nothing posts without your approval. Setup takes about two minutes: see set up your AI tool.