Is Ilmenite really a browser?

Yes — a headless browser engine built in pure Rust. It parses HTML, queries CSS selectors, runs JavaScript, and extracts content. Not a Chromium fork.

How does pricing work?

Credits-based, pay per use. 1 scrape = 1 credit. Free tier includes 500 credits/month. Pay as you go after that — no subscriptions, no browser-hour metering.

Can I self-host Ilmenite?

Self-hosting is available on the Enterprise plan. The engine ships as a single binary and a Docker image for VPC or air-gapped deployment.

newPDF extraction + Chrome JS rendering

Web scraping API
for AI agents.

Turn any URL into clean markdown, structured data, or extracted JSON. Scrape, crawl, map, extract, and search — one API, pay per use.

Read docs

0.19ms cold start
2MB per session
PDF + JS rendering
MCP for Claude

DEMOapi.ilmenite.dev/v1/scrape

Full playground →

Click Scrape to see an example response. Sign in to make live calls.

terminal

# Scrape any URL → clean markdown
curl -X POST https://api.ilmenite.dev/v1/scrape \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://news.ycombinator.com",
    "format": "markdown"
  }'

# → { "markdown": "# Hacker News\n\n1. Show HN: ..." }
# → 120ms total

api.ilmenite.devQuickstart →

0.19ms

Median cold start

2.4B

Pages parsed

99.98%

Uptime (90d)

47ms

p95 API latency

Head to head

Ilmenite vs Chrome headless.

Measured on Apple M2, release build, 100 sequential runs averaged. Chrome numbers from puppeteer launched with --headless=new.

Cold start

7368× better

Ilmenite

0.19ms

Chrome

1.4kms

RAM / instance

190× better

Ilmenite

2MB

Chrome

380MB

Binary size

15× better

Ilmenite

12MB

Chrome

180MB

Monthly cost (100 concurrent)

90× better

Ilmenite

Chrome

$450

Cost estimate: 100 concurrent sessions, 24h/day, on DigitalOcean · methodology

How it works

See every step your agent ran.

One API call can navigate, wait for selectors, click elements, and extract structured data — all in under a second.

news.ycombinator.com

session_7c9a8f

Hacker News

1. Show HN: A pure Rust browser engine (ilmenite.dev)

412 points by abhay · 3 hours ago · 187 comments

2. OpenAI releases o4-mini (openai.com)

298 points by sundeepr · 2 hours ago · 142 comments

3. Rust 1.85 released (rust-lang.org)

201 points by steveklabnik · 5 hours ago · 98 comments

Steps0.00s / 0.78s

navigatenews.ycombinator.com0.00s
wait_fortr.athing0.03s
page.markdown()4.2KB0.18s
clicka[href*=item?id]:nth(0)0.32s
wait_for.comment0.46s
extract(schema)32 comments0.61s
complete0.78s total0.78s

Primitives

Five endpoints. Everything your agent needs.

POST /v1/scrape

Turn any URL into clean markdown.

HTML → LLM-ready markdown, JSON, or structured text. JavaScript rendering, smart content extraction, adaptive parsing. The default call your agent will make 100× a day.

Read reference

bash

curl https://api.ilmenite.dev/v1/scrape \
  -H "Authorization: Bearer $KEY" \
  -d '{
    "url": "https://example.com",
    "formats": ["markdown", "links"]
  }'

POST /v1/crawl

Walk an entire site with one call.

Pass a seed URL, get every reachable page. Depth limits, URL globs, concurrency control, robots.txt support, webhook callbacks. 200 pages in under 2 seconds.

Read reference

typescript

const job = await client.crawl.start({
  url: "https://docs.stripe.com",
  maxPages: 200,
  include: ["/docs/*"],
});

for await (const page of client.crawl.stream(job.id)) {
  console.log(page.url);
}

POST /v1/extract

Schema in, typed data out.

Pass a JSON Schema, get validated structured data. Powered by OpenAI or Anthropic (bring your own key). Built for RAG pipelines, tool calls, and data enrichment.

Read reference

bash

curl https://api.ilmenite.dev/v1/extract \
  -H "Authorization: Bearer $KEY" \
  -d '{
    "url": "https://example.com/product/42",
    "schema": {
      "name": "string",
      "price_usd": "number",
      "in_stock": "boolean"
    },
    "provider": "openai"
  }'

# → { "data": { "name": "Widget", "price_usd": 29.99, "in_stock": true } }

POST /v1/map

Discover every URL on a site.

Pass a domain, get every discoverable URL. Parses sitemap.xml, follows links, filters by path. Feed the list into /crawl or /batch/scrape to extract them all.

Read reference

bash

curl https://api.ilmenite.dev/v1/map \
  -H "Authorization: Bearer $KEY" \
  -d '{
    "url": "https://docs.stripe.com",
    "limit": 500,
    "search": "/docs"
  }'

# → { "count": 342, "links": ["https://docs.stripe.com/...", ...] }

POST /v1/search

Google search + scrape in one call.

Send a query, get top results with snippets. Set scrape=true to also extract full markdown from each result page. Built for RAG and research agents.

Read reference

bash

curl https://api.ilmenite.dev/v1/search \
  -H "Authorization: Bearer $KEY" \
  -d '{
    "query": "rust web scraping 2026",
    "num_results": 5,
    "scrape": true
  }'

# → results with title, url, snippet, markdown

Cookbook

Recipes you can copy today.

Browse all recipes

Scrape an e-commerce catalog

Crawl product pages, extract typed records.

client.crawl.start({
  url: "shop.example.com",
  include: ["/products/*"],
  extract: ProductSchema,
});

Monitor price changes

Poll a URL, diff extracted fields, fire webhooks.

await client.extract({
  url: "flights.example.com/lax-jfk",
  schema: { price_usd: "number" },
  webhook_url: "https://my.app/hook",
});

Feed RAG with clean markdown

Walk docs site, chunk markdown, embed.

const pages = await client.crawl.stream(job);
for await (const p of pages) {
  await embedAndStore(p.markdown);
}

Give Claude a browser

Connect Ilmenite MCP to Claude Desktop.

// claude_desktop_config.json
{
  "mcpServers": {
    "ilmenite": { "url": "mcp.ilmenite.dev" }
  }
}

Stream sitemap → warehouse

Map a site, pipe URLs to BigQuery.

const urls = await client.map({
  url: "docs.example.com",
});
await bq.insert("pages", urls);

PDF → searchable text

Extract text from any PDF URL. Handles arxiv papers, invoices, scans.

const result = await fetch(API + "/v1/scrape", {
  method: "POST",
  body: JSON.stringify({
    url: "https://arxiv.org/pdf/2410.02073.pdf"
  })
});
const { content } = await result.json();
console.log(content.markdown);

Works with your stack

Drop into what you already have.

LangChain

Tool + DocumentLoader adapters. Python and TypeScript.

from ilmenite.langchain import IlmeniteScrapeTool

agent = initialize_agent(
    tools=[IlmeniteScrapeTool()],
    llm=ChatOpenAI(model="gpt-4o"),
)

Claude Desktop

Native MCP server.

{
  "ilmenite": {
    "url": "mcp.ilmenite.dev"
  }
}

Python

Sync + async, Pydantic extract.

pip install ilmenite

TypeScript

ESM, edge runtime, Zod extract.

npm install ilmenite

Rust

Cloud client or local engine. Same crate.

use ilmenite::Client;
let client = Client::from_env()?;
let md = client.scrape(url).send().await?.markdown;

Claude MCP

Give Claude web access. 5 tools: scrape, crawl, map, extract, search.

// claude_desktop_config.json
{
  "mcpServers": {
    "ilmenite": { "command": "ilmenite-mcp" }
  }
}

Compare

How Ilmenite stacks up.

Ilmenite is a managed API + self-hostable engine. Playwright is included as a reference point for the automation space.

Feature	Ilmenite	Browserbase	Steel.dev	PlaywrightOSS library
Cold start	0.19ms	500-2000ms	500-2000ms	500-2000ms
RAM per instance	~2 MB	200-500 MB	200-500 MB	200-500 MB
Single binary deploy
Chrome dependency	Optional	Required	Required	Required
Pure Rust
Native MCP server
Browser hours metered
Self-hostable

Pricing

Pay per use. Start free.

Credits-based pricing. 1 scrape = 1 credit. No subscriptions, no browser-hour metering.

Free

Try it out. No credit card.

500 credits / month

Start free

500 credits included
2 concurrent requests
All endpoints (scrape, crawl, map, extract, search)
PDF extraction
Community support

Developer

For builders shipping agents. No monthly commitment.

$0.001/ credit

Pay as you go

Get started

Pay only for what you use
10 concurrent requests
$5 min top-up
MCP server access
Chrome JS rendering
Email support

Best value

Pro

For production workloads at scale.

$0.0006/ credit

Volume pricing

Get started

40% cheaper per credit
50 concurrent requests
Priority queue
Webhook on crawl completion
99.9% uptime SLA
Slack support

Enterprise

For platforms and regulated industries.

Custom

Volume + self-host

Contact sales

Custom credit pricing
Unlimited concurrency
Self-hosted or air-gapped deploy
SSO, SAML, audit logs
99.99% SLA + dedicated engineer
SOC 2 Type II

Credit usage per operation

Scrape1 credit

Crawl (per page)1 credit

Map1 credit

Search2 credits

Extract (LLM)5 credits

PDF2 credits

Chrome JS render3 credits

FAQ

Questions, answered.

Ship your data pipeline in minutes.

Scrape, crawl, extract, search — one API. Your first 500 credits are free, no credit card required.

No credit card. 500 credits/month on the free tier. Pay as you go after.

Web scraping APIfor AI agents.

Ilmenite vs Chrome headless.

See every step your agent ran.

Five endpoints. Everything your agent needs.

Turn any URL into clean markdown.

Walk an entire site with one call.

Schema in, typed data out.

Discover every URL on a site.

Google search + scrape in one call.

Recipes you can copy today.

Scrape an e-commerce catalog

Monitor price changes

Feed RAG with clean markdown

Give Claude a browser

Stream sitemap → warehouse

PDF → searchable text

Drop into what you already have.

How Ilmenite stacks up.

Pay per use. Start free.

Free

Developer

Pro

Enterprise

Credit usage per operation

Questions, answered.

Is Ilmenite really a browser? It's only 2MB.

What about sites that need full Chrome rendering?

How does pricing work?

What's metered?

Can I self-host the engine?

Do you have SDKs?

Ship your data pipeline in minutes.

Web scraping API
for AI agents.