← All posts

May 22, 2026 · 4 min read

How to Set Up the AdWhispr MCP Server in Cursor

The complete AdWhispr MCP Cursor setup guide: connect the server, confirm all 8 tools, and run competitor ad research without leaving your editor.

You're building a landing page or wiring up ad code, and the question you actually need answered is "what's the competition running right now?" Tab-switching to a research tool breaks the flow. The AdWhispr MCP Cursor setup fixes that: it drops Meta competitor-ad intelligence directly into Cursor's chat, so you can pull a rival's longest-running hooks while your editor is still open on the file you're shipping.

This is a tight, practical walkthrough. Five minutes, mostly copy-paste.

What MCP gives you inside Cursor

MCP (Model Context Protocol) lets Cursor's AI call external tools. Connect AdWhispr's MCP server and Cursor gains read-only access to any brand's entire Meta ad library — ingested, snapshotted daily, and interrogable in plain English.

Concretely, while you build:

It is read-only on competitor data. AdWhispr never touches your live ad account and never launches anything — it reads Meta's public ad library and the daily history it builds on top.

Before you start

You need two things:

  1. Cursor with MCP support (Settings → MCP).
  2. An AdWhispr account. The free tier works to try this — $0, 1 brand, 5 messages/month. Enough to confirm the setup and run a few real queries before deciding on Pro.

Step 1: Add the AdWhispr server to Cursor

There are two paths. Pick one.

Option A — the config command (fastest)

Run this in your terminal:

npx adwhispr-mcp-server config

The package is unscoped — it's adwhispr-mcp-server, not @adwhispr/anything. The command walks you through authentication and writes the server entry into your Cursor MCP config for you. This is the recommended path because it handles the auth handshake without you hand-editing JSON.

Option B — add the server URL manually

If you'd rather configure it yourself, open Cursor's MCP settings and add a server pointing at:

https://adwhispr.com/api/mcp

Your mcp.json entry looks like this:

{
  "mcpServers": {
    "adwhispr": {
      "url": "https://adwhispr.com/api/mcp"
    }
  }
}

On first use you'll be sent through an OAuth flow to link your AdWhispr account. Approve it, and the connection sticks.

Step 2: Confirm the 8 tools appear

Restart Cursor (or reload the MCP server from Settings). Open the MCP panel and you should see AdWhispr listed with 8 tools:

Tool What it does
search_brands Find a brand already in the system
add_brand Ingest a new Facebook/brand URL's full ad library
get_brand_ads Pull a brand's ads, including days-running
get_brand_stats Summary stats for a brand's library
search_ads Semantic search across ads by hook, format, or theme
compare_brands Side-by-side on multiple competitors
clone_ad Generate original creative grounded in a real winner
generate_brief Export a competitive brief (PDF + Markdown)

If all eight show up green, you're done. If the server shows red or zero tools, re-run npx adwhispr-mcp-server config, check that your terminal completed the auth step, and reload the window.

Step 3: Run your first queries

Open Cursor's chat (the agent/composer pane) and just ask. A few that prove the setup end to end:

Ingest a competitor:

"Add Liquid Death to AdWhispr and give me its brand stats."

This calls add_brand, then get_brand_stats. First ingest takes a moment while it pulls the full library.

Find proven winners while you write copy:

"Show me Ridge Wallet's longest-running ads and the hooks they use."

This is the one you'll lean on. Days-running is AdWhispr's core performance proxy — an ad live for months is a winner Meta's API won't flag for you, because Meta returns no history. The history is the product. Read the distribution of run-times, not a single number.

Compare before you build:

"Compare Hims and Roman on ad formats and creative-iteration rate."

compare_brands shows you who's shipping new creatives fastest and where the format gaps are — useful context before you spec a competing page.

Generate a brief into your repo:

"Generate a competitive brief on Gymshark as Markdown."

generate_brief leads with the derived-intelligence panel — longevity curve, engagement-verified reach, iteration rate — then the qualitative read. Save the Markdown next to your project docs.

One honesty note that matters

You will not get a competitor's CTR, CPC, or ROAS from any of these tools — because Meta Ad Library doesn't expose them. Those numbers live only inside the advertiser's own account. Any tool that hands you a rival's exact ROAS invented it. AdWhispr instead cites the inputs behind every signal: spend and impressions come as wide ranges, and longevity, engagement, and iteration rate are derived from the daily snapshots it owns. That transparency is the point — build your decisions on signals you can actually verify.

When to move past the free tier

The free tier is built for exactly this: try the AdWhispr MCP Cursor setup, ingest one brand, run a handful of queries. When you need more brands tracked, unlimited tool calls, ad cloning, or full change alerts, Pro is $29/month (3 brands, 10 clones/mo, competitive briefs, 3-day free trial) and Agency is $149/month (10+ brands, 50 clones/mo, cross-brand comparison). Check adwhispr.com for current details.

That's the whole setup. The payoff: competitor ad intelligence one chat message away, without ever leaving the file you're shipping. For more on what the tools can do, browse the AdWhispr blog.

Wire it up once, and your next landing page starts with what's actually winning — not a guess.