Bring your bill.
If you're building agents, RAG pipelines, or AI systems that turn web content into structured data, there's a real chance Gyrence has lower total cost of ownership (TCO) than what you're paying now.
If we don't, we'll tell you that.
Send us your last invoice from any provider — HTML scrapers, AI extraction services, all of them. We'll respond within one business day with an honest TCO comparison for your actual workload — invoice, plus the engineering time you're spending to keep your current setup working.
Who we're for
If you're building agents that browse and reason over web content, RAG systems pulling fresh data into LLM prompts, training pipelines that need structured records at scale, or any product where the end output is clean JSON your AI consumes — keep reading. Gyrence's primary surface is HTML pages: we fetch them (with JavaScript rendering when needed), clean them, and extract structured data via our AI extraction layer. For other formats — JSON APIs, RSS feeds, PDFs, Office documents — see the support matrix below.
Most providers stop at one format. HTML scrapers don't handle your markdown documentation. Markdown processors don't crawl. You end up integrating multiple tools, paying multiple bills, and writing the glue code to stitch them together. The invoice looks cheap. The real cost shows up in engineering hours — selectors that break, parsers your team writes by hand, NLP for unstructured fields, and the integration tax of running a stack of disconnected services.
Gyrence's primitives — Search, Gyre, Fetch, Extract, Map — share the same credit pool, same workspace, same bill. The Extract layer takes a target schema and returns structured JSON best-effort matching that schema, regardless of source site's HTML structure. One service, one integration, one set of credits across all five operations. Your TCO is the bill — not the bill plus the people-time to keep a stack of disconnected services glued together.
Building a financial-news intelligence product
Your product reads earnings coverage, market commentary, and corporate-action news from Reuters, Bloomberg, Yahoo Finance, MarketWatch, and a long tail of financial publishers to surface trends and signals for an AI analyst. You process around 50,000 article pages a month.
What you actually want from each article:
{
"publisher": "Reuters",
"headline": "Apple Q4 revenue beats estimates on iPhone strength",
"published_iso": "2026-05-28T14:22:01Z",
"primary_ticker": "AAPL",
"mentioned_companies": ["AAPL", "MSFT", "GOOGL"],
"sentiment": "positive",
"key_facts": [
"Q4 revenue of $94.9B vs. $94.4B expected",
"iPhone revenue up 6% YoY",
"Services hit record $24.2B"
],
"summary": "Apple beat Q4 expectations driven by iPhone..."
}Side-by-side TCO:
| HTML-first path | GYRENCE | |
|---|---|---|
| Monthly invoice | $50 – $150 | $92.50 |
| Engineering tax (10–20 hrs/mo × $150 loaded) | $1,500 – $3,000 | $0 |
| Total monthly TCO | $1,550 – $3,150 | $92.50 |
Each publisher has different page structures, paywalls handled differently, layouts that change without notice. Your team writes selectors for each major source, maintains them as sites update, builds NLP for sentiment and key-fact extraction, then handles the long tail of smaller publishers individually. Gyrence's Extract layer takes a target schema and returns structured JSON best-effort matching it, regardless of source site's HTML structure. New publishers don't require new parsers; layout changes rarely break extraction because the schema is what's specified, not the path through the DOM. For very long pages, the first ~12,000 characters are processed — Gyre + Extract together handle coverage when content spans pages.
Building a developer-tool intelligence product on GitHub repos
Your product tracks open-source projects to surface trending tools, dependency adoption, and contributor patterns for a developer analytics dashboard. You process around 80,000 repo pages a month.
What you actually want from each repo:
{
"name": "anthropic-sdk-python",
"owner": "anthropics",
"stars": 4823,
"primary_language": "Python",
"license": "MIT",
"last_commit_iso": "2026-05-28T14:22:01Z",
"top_5_contributors": ["alice", "bob", "carol", "dave", "eve"],
"has_security_policy": true,
"open_issues_count": 47,
"readme_summary": "Official Python SDK for the Anthropic API..."
}Side-by-side TCO:
| HTML-first path | GYRENCE | |
|---|---|---|
| Monthly invoice | $30 – $80 | $142.50 |
| Engineering tax (5–10 hrs/mo × $150 loaded) | $750 – $1,500 | $0 |
| Total monthly TCO | $780 – $1,580 | $142.50 |
GitHub changes their markup roughly every quarter. Each change breaks parsers and triggers patching work. The public API is heavily rate-limited and doesn't expose every field you need. Gyrence's Extract layer adapts to markup changes because you specify the schema you want, not the path through the DOM.
Building a media-analytics product on IMDb
Your product surfaces film and TV intelligence — ratings trends, cast networks, release-date analytics — for a recommendation engine or media-research dataset. You process around 120,000 title pages a month.
What you actually want from each title:
{
"title": "Dune: Part Two",
"year": 2024,
"imdb_rating": 8.5,
"runtime_minutes": 166,
"director": "Denis Villeneuve",
"top_5_cast": ["Timothée Chalamet", "Zendaya", "Rebecca Ferguson", "Javier Bardem", "Josh Brolin"],
"genre_tags": ["Action", "Adventure", "Drama", "Sci-Fi"],
"plot_summary": "Paul Atreides unites with Chani and the Fremen..."
}Side-by-side TCO:
| HTML-first path | GYRENCE | |
|---|---|---|
| Monthly invoice (proxy + rendering) | $120 – $200 | $202.50 |
| Engineering tax (8–15 hrs/mo × $150 loaded) | $1,200 – $2,250 | $0 |
| Total monthly TCO | $1,320 – $2,450 | $202.50 |
IMDb's pages are JavaScript-heavy with anti-bot defenses, so HTML-first paths need rendering plus proxy rotation. They restructured the cast section twice in 2025. Filmography layout changes happen on a rolling basis. Gyrence renders JavaScript-heavy pages via our self-hosted browser fleet and extracts structured data via the same primitive — when IMDb changes layout, your code doesn't. Common bot-defense challenges (Cloudflare, Akamai, PerimeterX, DataDome) are detected and surfaced as honest errors rather than saving challenge pages as 'content'. If a site deploys an anti-bot system we can't get through, you'll know immediately rather than silently.
Worked example: an AI builder processing 80,000 pages a month
40K bare scrapes for monitoring, 40K with AI extraction for ingestion. The current path uses an HTML provider plus a separate AI-extraction subscription.
Side-by-side monthly invoice:
| Current path | GYRENCE | |
|---|---|---|
| Base provider plan (50K credits) | $99 | — |
| AI extraction subscription | $199 | — |
| Overage at typical rates | ~$30 | — |
| Gyrence PAYG (80K credits, graduated rates) | — | $137.50 |
| Gyrence base | — | $5.00 |
| Total monthly invoice | $328 | $142.50 |
Same primitives. Same volume. Extract bundled. 57% lower invoice. Fully predictable. No second subscription. And on TCO, the gap widens further — the current path's $328 doesn't include the engineering time spent integrating two services and reconciling two bills.
Note: PAYG charges 1 credit for HTTP-only fetches and 3 credits for browser-rendered fetches. The 80K-credit estimate above assumes a typical mix; your actual mix depends on which sites you target. See /pricing for a more specific estimate.
What works today, what doesn't
We'd rather you know up front than discover it after you start.
Works today:
- HTML pages, including JavaScript-rendered SPAs (via our self-hosted browser fleet)
- Plain-text content served as HTML
- Sitemap XML (for site mapping via the Map primitive)
- Search-engine results (via the Search primitive, powered by Brave Search)
- AI extraction from the above (via the Extract primitive)
Doesn't work today:
- PDFs and Office documents (Word, Excel, PowerPoint)
- JSON, XML, CSV, RSS, and other structured-data formats served at a URL — the Extract primitive expects HTML
- Images, audio, and video
- Sites protected by Akamai Bot Manager — we don't bypass it, and our browser fleet can't either
If your workload is centered on the second list, we're not the right fit yet. Email us anyway — we'll tell you honestly which side of this line you're on, and where you'd find the better fit.
Where we don't win
A few more honest disqualifications. We'd rather you know now than discover it three months in.
- If your team prefers to own the parsing layer — you have engineering capacity, you've already built scrapers for your target sites, and you'd rather pay engineering time than provider fees — there are HTML providers that beat us on invoice cost. We don't compete on raw-HTML-per-dollar. We compete on TCO for teams that would rather their engineers ship features than maintain selectors.
- If your monthly spend is under $50, our $5 base plus PAYG rates probably lose to entry-tier providers. Gyrence isn't built for hobbyists or very-low-volume usage.
- If your target sites are well-served by pre-built scrapers — common e-commerce sites, major search engines, well-known profile platforms — some providers maintain libraries of ready-made scrapers for those specific targets. Use them for that. Use us for the long tail of sites that don't have a pre-built option.
- If your workload is dominated by enterprise proxy needs — large residential or mobile pools, geo-distributed sourcing for ad verification or geo-testing — there are providers with structurally bigger networks. They built something different than we did.
- If your target sites are protected by Akamai Bot Manager — American Airlines, parts of LinkedIn, many enterprise SaaS sites, some financial services platforms — we don't bypass it today. Some specialty providers handle Akamai specifically. We'll be honest if your target list is Akamai-heavy and recommend you stay where you are for those sites.
Send us your bill anyway. We'll tell you honestly which side of this line you're on.
What we offer if your workload fits
Send us your last month's invoice. We'll grant 2× the spend amount as Gyrence credits, burnable over 90 days. Try us at no incremental cost.
Our team rewrites your existing scrape, crawl, and extract calls to Gyrence's primitives — Search, Gyre, Fetch, Extract, Map. Usually a day, not a sprint.
Run Gyrence alongside your current provider for a month using the credit grant. We'll send you a weekly comparison report (sent by our team during your parallel run) showing freshness, success rate, and cost differences. Make the cutover decision with real data.
What moves with you
- Five primitives: Search, Gyre, Fetch, Extract, Map — same credit pool, same workspace
- Every fetch result carries the source URL, fetch timestamp, and the tier used (HTTP or browser rendering) — visible in the response, captured in your usage log
- Common bot defenses (Cloudflare, Akamai, PerimeterX, DataDome) are detected and surfaced as honest errors rather than saving challenge pages as 'content'
- MCP endpoint included — point Claude, Cursor, or any MCP client at your workspace key. Same billing as HTTP
- OpenAPI spec at /openapi.json — describes the same Zod schemas used by the HTTP routes
- Predictable monthly bill — spending caps you set, hard credit limits per tier, usage warnings as you approach those limits
Send us your invoice
Email a PDF, screenshot, or pasted line items to switch@gyrence.com. Include:
- Your current provider
- Approximate monthly spend
- What you're using the API for (scraping for monitoring? extracting for AI? both?)
We'll reply within one business day with:
- An honest TCO comparison for your specific workload — provider invoice plus the engineering time you're spending today
- A like-for-like primitive mapping (your provider's calls → ours)
- A sandbox workspace pre-loaded with your 2× credit grant
- A 30-day parallel-run plan, if it makes sense for you