x-mcp
Read-only MCP server for Kan's FxTwitter API proxy, enabling public X/Twitter search and retrieval.
README
x-mcp
Read-only MCP server for public X/Twitter search and retrieval via twitter.2-38.com. Implemented as a TypeScript + Hono Cloudflare Worker.
日本語版は下にあります: 日本語
Remote MCP endpoint
https://x.mcp.2-38.com/mcp
Discovery metadata:
https://x.mcp.2-38.com/.well-known/mcp.json
This server exposes MCP tools that call the existing public proxy endpoints under:
https://twitter.2-38.com/api/fx/2/...
Example upstream call wrapped by search_posts:
https://twitter.2-38.com/api/fx/2/search?q=from%3Ajack&feed=latest&count=10
AI agent skill first
MCP tells an agent what tools exist, but a skill tells it when and how to use them. Before asking an AI agent to use this MCP, have it install or load the bundled skill:
https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md
Full copy/paste prompts are available here:
Short English prompt:
Install or load this x-mcp skill first:
https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md
Then configure MCP server `x` with:
https://x.mcp.2-38.com/mcp
Verify tools/list includes search_posts and call search_posts with
{ "q": "from:jack", "feed": "latest", "count": 2 }.
Use this instead of x_search/xurl for public read-only X/Twitter search and retrieval through twitter.2-38.com.
MCP configuration
Hermes Agent
hermes skills inspect https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md
hermes skills install https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md --category social-media -y
hermes mcp add x --url https://x.mcp.2-38.com/mcp
hermes mcp test x
Or add this to ~/.hermes/config.yaml and restart Hermes / start a new session:
mcp_servers:
x:
url: "https://x.mcp.2-38.com/mcp"
Hermes will expose tools with the mcp_x_ prefix, for example mcp_x_search_posts.
Generic MCP clients
{
"mcpServers": {
"x": {
"url": "https://x.mcp.2-38.com/mcp"
}
}
}
Use your client's HTTP / Streamable HTTP MCP transport setting if it requires an explicit transport type.
Tools
search_posts— call/2/searchget_post— call/2/status/:idget_profile— call/2/profile/:handleget_profile_statuses— call/2/profile/:handle/statusesget_profile_media— call/2/profile/:handle/mediaget_trends— call/2/trendstypeahead— call/2/typeaheadget_openapi— call/2/openapi.json
Smoke test
curl -sS https://x.mcp.2-38.com/health
curl -sS https://x.mcp.2-38.com/.well-known/mcp.json
For an MCP-level smoke test, call tools/list and then search_posts using your MCP client. A successful search_posts call with q=from:jack returns JSON text containing:
{
"upstreamUrl": "https://twitter.2-38.com/api/fx/2/search?q=from%3Ajack&feed=latest&count=2",
"status": 200
}
Local development
npm install
cp wrangler.example.jsonc wrangler.jsonc
npm run check
npx wrangler dev --config wrangler.jsonc
npm run check regenerates local Worker runtime types from wrangler.example.jsonc and runs TypeScript type checking.
MCP endpoint locally:
http://localhost:8787/mcp
Deploy
cp wrangler.example.jsonc wrangler.jsonc
# Fill account_id in wrangler.jsonc if needed.
npm run deploy
wrangler.jsonc is intentionally gitignored so account-specific deployment config stays local.
Security and scope
- Public, unauthenticated, read-only MCP endpoint.
- Do not include secrets, private data, or access tokens in search queries.
- Not for posting, liking, reposting, following, DMs, notifications, private accounts, or authenticated timelines.
- Upstream availability, rate limits, and response shape depend on
twitter.2-38.com.
日本語
twitter.2-38.com 経由で公開X/Twitter情報を read-only で検索・取得する MCP server です。実装は TypeScript + Hono の Cloudflare Worker です。
公開MCP endpoint
https://x.mcp.2-38.com/mcp
Discovery metadata:
https://x.mcp.2-38.com/.well-known/mcp.json
このMCPは、既存の公開proxy endpointを呼びます。
https://twitter.2-38.com/api/fx/2/...
search_posts が包む upstream 呼び出し例:
https://twitter.2-38.com/api/fx/2/search?q=from%3Ajack&feed=latest&count=10
まずAIエージェントにskillを入れる
MCPだけだと「どの場面で、どう使うべきか」がAIに伝わりにくいので、このrepoにはAIエージェント向けskillを同梱しています。
Skill URL:
https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md
コピペ用の導入プロンプトはこちらです。
短い日本語プロンプト:
まず以下の x-mcp skill をインストール、または読み込んでください。
https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md
その後、MCP server `x` を以下で設定してください。
https://x.mcp.2-38.com/mcp
`tools/list` に `search_posts` が出ることを確認し、
`search_posts` を { "q": "from:jack", "feed": "latest", "count": 2 } で試してください。
このMCPは twitter.2-38.com 経由の公開X/Twitter情報取得専用です。公開read-only用途では x_search/xurl ではなくこのMCPを使ってください。
MCP設定
Hermes Agent
hermes skills inspect https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md
hermes skills install https://raw.githubusercontent.com/kandotrun/x-mcp/main/skills/x-mcp/SKILL.md --category social-media -y
hermes mcp add x --url https://x.mcp.2-38.com/mcp
hermes mcp test x
CLIが使えない場合は、~/.hermes/config.yaml に次を追加してHermesを再起動、または新しいセッションを開始してください。
mcp_servers:
x:
url: "https://x.mcp.2-38.com/mcp"
Hermesでは mcp_x_search_posts のように mcp_x_ prefix付きでtoolが見えます。
汎用MCP client
{
"mcpServers": {
"x": {
"url": "https://x.mcp.2-38.com/mcp"
}
}
}
client側でtransport指定が必要な場合は、HTTP / Streamable HTTP MCP transportを選んでください。
Tools
search_posts—/2/searchを呼ぶget_post—/2/status/:idを呼ぶget_profile—/2/profile/:handleを呼ぶget_profile_statuses—/2/profile/:handle/statusesを呼ぶget_profile_media—/2/profile/:handle/mediaを呼ぶget_trends—/2/trendsを呼ぶtypeahead—/2/typeaheadを呼ぶget_openapi—/2/openapi.jsonを呼ぶ
Smoke test
curl -sS https://x.mcp.2-38.com/health
curl -sS https://x.mcp.2-38.com/.well-known/mcp.json
MCPレベルでは、clientから tools/list → search_posts を確認してください。q=from:jack の成功例では、返ってくるJSON textに次のような内容が含まれます。
{
"upstreamUrl": "https://twitter.2-38.com/api/fx/2/search?q=from%3Ajack&feed=latest&count=2",
"status": 200
}
ローカル開発
npm install
cp wrangler.example.jsonc wrangler.jsonc
npm run check
npx wrangler dev --config wrangler.jsonc
npm run check は wrangler.example.jsonc からローカルのWorker runtime型を再生成し、TypeScriptの型チェックを実行します。
ローカルMCP endpoint:
http://localhost:8787/mcp
デプロイ
cp wrangler.example.jsonc wrangler.jsonc
# 必要に応じて wrangler.jsonc の account_id を埋める
npm run deploy
wrangler.jsonc はアカウント固有の設定なのでgitignoreしています。
セキュリティと範囲
- 公開・認証なし・read-only のMCP endpointです。
- search query に secret / private data / access token を入れないでください。
- 投稿、いいね、RT、フォロー、DM、通知、private account、認証timelineには使えません。
- upstreamの可用性、rate limit、response shapeは
twitter.2-38.com側に依存します。
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.