If your agent runs in an MCP client — Claude Desktop, Claude Code, Cursor, Cline, Windsurf — it can answer "how did this post do?" with real numbers from a link. The whole setup is one JSON block.
{
"mcpServers": {
"pulse": { "command": "npx", "args": ["-y", "pulse-mcp"] }
}
}
That's everything. No account, no API key, no environment variables. On first run
npx fetches the package and you'll see
[pulse-mcp] ready — tools: metrics, metrics_batch, history, profile on stderr.
(Node ≥ 20. Also on the official MCP registry
as io.github.patwalls/pulse-mcp.)
metrics — paste any public post URL (YouTube, X/Twitter incl. view
counts, TikTok, Bluesky, Mastodon, Instagram) and get
{ views, likes, comments, shares, quotes, bookmarks, publishedAt, title, author, thumbnail }.
Short links (vm.tiktok.com, t.co) resolve automatically.
metrics_batch — up to 50 URLs in one call, posts and profiles mixed, order preserved, one bad URL never fails the batch.
history — the growth curve: every fresh read records a snapshot, and
since returns only the new points. Ask "is this post still growing?" and the agent
answers from the slope.
profile — a profile URL → followers, following, post count, verified.
Ask your agent: "How many views does
https://www.youtube.com/watch?v=dQw4w9WgXcQ have?" — it should call metrics and
answer with a real number north of 1.78 billion.
The server is a thin shim over the free public Pulse API,
which reads each platform's own public pages from a residential IP — so it sees what a real browser
sees where datacenter IPs get walled. Honest errors: deleted posts are
content_unavailable, login-walled platforms say login_required, never fake
zeros. Free for individuals; commercial/volume use has a $19/mo Pro plan (higher rate limit).
More: the blog ·
API docs ·
the server's repo (includes
llms-install.md for agent-driven setup).