Live X (Twitter) data as native tools for Claude, Cursor, and any MCP client. One command to install, read-only, no X developer account.
MCP is the open standard that lets an AI client call external tools in one uniform way. The Sorsa API MCP server is the X (Twitter) piece of that: it wraps Sorsa API's read-only data endpoints so any MCP-compatible client gains tools for searching tweets and reading profiles, followers, replies, mentions, trends, and more. You ask your AI in plain language, it picks the right tool and calls it. It is read-only by design: no posting, no logins, no scraping.
Every tool maps to a Sorsa API endpoint and returns public X data only. The model picks the right one for each request.
| Tool | What it returns | Typical use |
|---|---|---|
| search_tweets | Advanced tweet search with the full operator set (from:, since:, lang:, min_faves:). | Topic monitoring, dataset building |
| search_users | Accounts matching a keyword or phrase. | Audience and lead discovery |
| get_user_profile | Full profile by handle or ID: bio, follower counts, verification. | Account lookup and enrichment |
| get_user_followers | An account's followers with metadata, paginated. | Audience and network analysis |
| get_user_following | The accounts a user follows. | Interest-graph mapping |
| get_user_tweets | Recent tweets from a user's timeline. | Activity tracking and summaries |
| get_tweet | Full data for one or many tweets by URL or ID, author included. | Hydrating tweets with metrics |
| get_tweet_comments | The reply thread under a tweet. | Conversation analysis |
| get_tweet_quotes | Quote tweets of a given tweet. | Spread and discourse analysis |
| get_tweet_retweeters | The accounts that retweeted a tweet. | Amplification and reach |
| get_mentions | Tweets mentioning an account, with like, reply, and date filters. | Brand and reputation monitoring |
| get_trends | Trending topics for a location, by WOEID. | Trend detection and timely content |
| verify_action | Check if a user followed, retweeted, commented, quoted, or joined a community. | Campaign and giveaway verification |
Plus list and community tools. See the full endpoint set in the docs.
The server runs locally through npx, so there is nothing to clone or build. Add the config for your client, set your Sorsa API key, and restart it.
One command from your terminal, nothing to paste into a config file.
claude mcp add --transport stdio --env SORSA_API_KEY=YOUR_API_KEY sorsa -- npx -y @sorsa/mcp
Add this to claude_desktop_config.json (Settings, Developer, Edit Config), then restart Claude.
{ "mcpServers": { "sorsa": { "command": "npx", "args": ["-y", "@sorsa/mcp"], "env": { "SORSA_API_KEY": "YOUR_API_KEY" } } } }
Settings, Tools and Integrations, New MCP Server, or paste this into mcp.json.
{ "mcpServers": { "sorsa": { "command": "npx", "args": ["-y", "@sorsa/mcp"], "env": { "SORSA_API_KEY": "YOUR_API_KEY" } } } }
Add this to your MCP config (settings.json). Note the key is "servers", not "mcpServers".
{ "servers": { "sorsa": { "command": "npx", "args": ["-y", "@sorsa/mcp"], "env": { "SORSA_API_KEY": "YOUR_API_KEY" } } } }
Add the same mcpServers block in your Windsurf MCP settings (mcp_config.json).
{ "mcpServers": { "sorsa": { "command": "npx", "args": ["-y", "@sorsa/mcp"], "env": { "SORSA_API_KEY": "YOUR_API_KEY" } } } }
Every client uses the same standard block, just bring your Sorsa API key and restart the client after adding it.
The server authenticates with your own Sorsa API key, passed through the SORSA_API_KEY environment variable in the config above. Sign in, copy your key, paste it in, and restart your client.
Sign in at Sorsa API and open your dashboard.
Copy your API key.
Paste it as SORSA_API_KEY and restart your client.
With the server connected you never call the API directly. You ask in plain language, the client picks the tool, calls it, and reads the result back.
You ask, in Claude or Cursor, in plain language: "Get the profile of @naval."
The client picks the tool and calls it.
{ "name": "get_user_profile", "arguments": { "username": "naval" } }
The server returns live data.
{ "username": "naval", "display_name": "Naval", "followers_count": 2600000, "followings_count": 800, "verified": true, "created_at": "2007-04-09T17:51:23Z" }
The client answers in plain language, a table, or whatever you asked for. No glue code and no rate-limit handling on your side.
The same pattern covers every tool. "What is trending in the US" routes to get_trends, "recent tweets from @sama about AI" routes to search_tweets.
The server is only as good as the API behind it.
No X approval, no app review, just a key.
1 request equals 1 call on every endpoint and plan, so agent loops stay predictable.
The same rate limit on every plan, so your client is not throttled.
Tools can read public X data but never post, delete, or follow.
An MCP server that gives Claude, Cursor, and any MCP client read-only tools for live X (Twitter) data: search, profiles, followers, tweets, mentions, trends, and verification. Install it once and your AI calls these tools by itself.
Any MCP-compatible client: Claude Desktop, Claude Code, Cursor, VS Code with GitHub Copilot, Windsurf, and others. The setup is the same standard config in each.
The server is open-source on GitHub. It calls the Sorsa API, so you bring your own Sorsa API key and pay only for the requests the tools make.
Yes. The server reads your key from the SORSA_API_KEY environment variable. Creating a key takes a couple of minutes in the dashboard.
No. Every tool is read-only. The server reads public X data but never posts, replies, likes, follows, or sends DMs, which keeps autonomous agents safe.
The official X API needs a developer account, gates access behind expensive tiers, and is not built for MCP. Sorsa API needs no dev account, uses flat per-request pricing, and ships as native MCP tools.
No. The config launches the server through npx, which fetches the latest version automatically. You set your API key and restart your client.
Paste the Sorsa API docs URL into your LLM and it writes the integration code for you. See Build with AI.
Install the server, drop in your key, and Claude or Cursor can search tweets and read profiles, followers, and trends in the next message you send.
Prefer to generate code instead of running a server? Paste our docs into your LLM.
Build with AI