CaptureKit
Guides pratiques

Analyze a Page with AI

Send a URL to the CaptureKit AI analysis endpoint and receive structured insights about the page's content, design, technology stack, and SEO metadata.

Overview

The CaptureKit AI analysis endpoint combines screenshot capture with LLM-powered analysis. Pass any URL and a custom prompt — the API returns a structured JSON answer grounded in what the page actually looks like. Use it for competitor audits, UX reviews, or automated content QA.

Prerequisites

  • A CaptureKit API key — get one at app.capturekit.dev
  • Install dependencies for your language:
pip install requests

Steps

Submit an analysis request

Call GET /v1/analyze with the url and prompt parameters. The prompt shapes the AI's response.

import requests

API_KEY = "YOUR_API_KEY"
PROMPT  = "Summarize the main value proposition of this page in 2 sentences. Then list the top 3 CTAs visible above the fold."

response = requests.get(
    "https://api.capturekit.dev/v1/analyze",
    headers={"x-api-key": API_KEY},
    params={"url": "https://stripe.com", "prompt": PROMPT},
)
data = response.json()
print(data)

Read the AI analysis

The response includes analysis (the AI's answer to your prompt), screenshot_url, and metadata like title and description.

print(f"Page    : {data.get('title')}")
print(f"Preview : {data.get('screenshot_url')}")
print()
print("=== AI Analysis ===")
print(data.get("analysis", ""))

Run a competitor audit across multiple pages

Analyze several competitor landing pages with a consistent prompt and save the results as a structured report.

import json, time, requests

API_KEY = "YOUR_API_KEY"
PROMPT  = """Analyze this landing page and return a JSON object with these keys:
- value_proposition (string)
- primary_cta (string)
- target_audience (string)
- pricing_visible (boolean)
- social_proof_types (array of strings: "testimonials", "logos", "stats", etc.)"""

COMPETITORS = [
    {"name": "Stripe",  "url": "https://stripe.com"},
    {"name": "Paddle",  "url": "https://paddle.com"},
    {"name": "Lemonsqueezy", "url": "https://lemonsqueezy.com"},
]
report = []

for comp in COMPETITORS:
    r    = requests.get("https://api.capturekit.dev/v1/analyze",
        headers={"x-api-key": API_KEY},
        params={"url": comp["url"], "prompt": PROMPT})
    data = r.json()
    analysis_text = data.get("analysis", "{}")
    try:
        # AI may return the JSON wrapped in markdown fences
        raw = analysis_text.strip().removeprefix("```json").removesuffix("```").strip()
        analysis = json.loads(raw)
    except json.JSONDecodeError:
        analysis = {"raw": analysis_text}

    report.append({"competitor": comp["name"], "url": comp["url"], **analysis})
    print(f"✓ {comp['name']} analyzed")
    time.sleep(2)

with open("competitor_audit.json", "w") as f:
    json.dump(report, f, indent=2)
print("\nReport saved to competitor_audit.json")

Instruct the AI to return structured JSON in your prompt — the model will format its output accordingly, making it easy to parse and store results in a database or spreadsheet without additional post-processing.

Sur cette page