API · MCP server · built for agents

Competitor ad intelligence your agents can call

Divisadero turns any Meta or TikTok advertiser's live ads into structured, scored, queryable data. Reach it as a REST API or a native MCP server, so Claude and your own agents can track and analyze the competition in a single call.

Request API access Explore the live data

example · your agent using the Divisadero MCP server

You

Which sleep angles is AG1 running on Meta and TikTok that Ritual isn't?

Agent

tool call

divisadero.find_whitespace({ advertisers: ["ag1", "ritual"], platforms: ["meta", "tiktok"], axis: "angle" })

AG1 runs 3 sleep angles Ritual doesn't:

cortisol / stress-sleep18 adshook 84
travel & shift-work recovery9 adshook 79
magnesium-for-sleep6 adshook 82

Illustrative example. Endpoints and keys are issued on request.

How it works

01
Point it at an advertiser
Any brand or Page on Meta or TikTok, by name.
02
We pull the live ads
Straight from the platforms' public ad libraries, on demand.
03
Structure + score each
One fixed schema: hook, format, persona, claims, 0-100 scores.
04
Serve it to code or agents
Clean JSON over REST, or native tools over MCP.

Any advertiser

on Meta & TikTok

Every ad

parsed into one schema

Scored 0-100

hook · persuasion · craft

MCP-native

drop into any agent

What you can pull

The public ad libraries are searchable but raw. This is the analysis layer on top of them, built to be called by code and by agents. It is the same method we lay out, step by step, in our free guide to analyzing ads with AI, turned into an endpoint.

Track any advertiser

Point it at any brand or Page on Meta or TikTok. We pull their live ads from the public ad libraries, not just a fixed list of brands.

Structured creative

Every ad parsed into one fixed schema: hook, format, offer, on-screen claims, implied persona, production style. Comparable across thousands of ads and both platforms.

Creative scores

Hook, persuasion, execution, and credibility, each 0-100 on a fixed rubric, so ads are sortable and craft-vs-credibility gaps are visible.

Whitespace & gaps

The format x persona and angle combinations a competitor is not running, per platform, surfaced as first-class results.

Change tracking

New ads, killed ads, and shifts in a competitor's format and angle mix over time. Subscribe to an advertiser and get a webhook when it moves.

Compare & cross-platform

Diff two or more advertisers' playbooks in one call, or diff a single brand's Meta vs TikTok strategy. Who owns which angle, who is late to a format.

Two ways to call it

REST API

Clean JSON over HTTPS. Authenticate with a bearer key, ask for any advertiser, get structured, scored ads back.

example · bash

curl https://api.divisadero.co/v1/advertisers/ag1/ads?structured=true \
  -H "Authorization: Bearer sk_live_..."

{
  "advertiser": "AG1",
  "platform": "meta",
  "active_ads": 49,
  "ads": [
    {
      "id": "968792926318549",
      "format": "Problem-Solution",
      "hook": "Tired at 3pm? It might be your coffee.",
      "persona": "busy professional",
      "scores": { "hook": 86, "persuasion": 92, "credibility": 71 },
      "running_days": 214
    }
  ]
}

GET/v1/advertisers?q=find an advertiser by name
GET/v1/advertisers/{id}/adslist an advertiser's ads, structured + scored
GET/v1/ads/{id}one ad, full analysis
GET/v1/advertisers/{id}/whitespacethe angles and combos it isn't running
POST/v1/comparediff advertisers, or a brand across platforms
POST/v1/trackwatch an advertiser, webhook on change

MCP, for agents

Add one server to Claude or your own agent and the ad-intelligence tools show up natively. No glue code, no wrappers.

example · mcp config

{
  "mcpServers": {
    "divisadero": {
      "url": "https://mcp.divisadero.co",
      "headers": { "Authorization": "Bearer sk_live_..." }
    }
  }
}

tools your agent gets

search_advertisersget_adsanalyze_creativecompare_advertisersfind_whitespacetrack_advertiser

What people build

Competitive-monitoring agents

An agent that watches your rivals across Meta and TikTok and reports every new angle and format it sees each week, with the ads and scores attached.

Creative-research copilots

Pull the category's highest-scoring hooks and the whitespace before a shoot, so the brief is grounded in what's actually running.

Launch alerting

Get a webhook the moment a competitor ships a new format, angle, or offer, on either platform, instead of finding out a quarter later.

Your own dashboards

Feed structured, scored ad data straight into your BI or internal tools. The taxonomy is fixed, so the joins are clean.

What this data is, and isn't

Everything comes from the public Meta and TikTok ad libraries, so you get creative and runtime, not spend, impressions, CTR, or ROAS. The scores are directional creative-quality reads on a fixed rubric, not performance metrics, and runtime is the one real performance proxy (brands kill losers). The API tells you where to look. A test tells you if it's real. We would rather say that plainly than sell you a number we can't see.

API & MCP: FAQ

What data does the API return?

For any advertiser, its live and recent Meta and TikTok ads turned into structured data: the hook, creative format, offer, on-screen claims, implied persona, and production style, plus 0-100 creative scores and how long each ad has been running. Everything is sourced from the platforms' public ad libraries.

Which platforms does it cover?

Meta (Facebook and Instagram) and TikTok today. Both are pulled from the public ad transparency libraries and run through the same structuring and scoring, so a brand's Meta and TikTok creative is directly comparable in one schema.

Can I track any advertiser, or only brands you already list?

Any advertiser running on Meta or TikTok. Point us at a brand name or a Page and we pull and analyze its ads on demand. The brands on divisadero.co/explore are just a public sample of what the pipeline does.

What is the MCP server?

A Model Context Protocol server. Add it to Claude or any MCP-compatible agent and the agent can call the ad-intelligence tools directly (search advertisers, get ads, compare, find whitespace) with no glue code. The same data as the REST API, exposed as native agent tools.

What are the honest limits?

The public ad libraries show creative and runtime, not spend, impressions, CTR, or ROAS. So the scores are directional creative-quality reads on a fixed rubric, not performance metrics, and runtime is the one real performance proxy (brands kill losers). The API tells you where to look. A test tells you if it's real.

How do I get access and what does it cost?

Access is early and issued by request. Pricing scales with how many advertisers you track and how many calls you make. Request a key and we'll get you set up.

Point an agent at your competitors

Request a key and start pulling structured, scored Meta and TikTok ad data, from your code or straight from your agent.

Request an API key See the data first

New here? See the Wellness Ad Index or read the method.