GeezerVison

The World Wide Web, restored to 800 pixels of glory.

GeezerVison is showing spider.cloud — AI Web Crawler API for Agents and RAG | Spider

Infrastructure for reading the public web.

One API to scrape, crawl, and search, backed by our own browser engine and network. Low cost, low latency, built to scale.

Get started free

Run requests before you buy. Or paste the setup prompt into your coding agent.

api.spider.cloud POST
100K+ pages per second
Crawl, scrape and search on one host. Whole sites stream back as markdown, in order.
browser.spider.cloud WS
100% browser fidelity bench v1
Spider, 80 tasks, Sep 18, 2026. Real production sites. See the results →
proxy.spider.cloud HTTP
215M+ exits in 199 countries
Residential addresses, rotated per request. The same network runs the crawler and the browser.

Integrations→

Scraping API, browser, proxy. One key.

Start with the API. The browser and the proxies each work on their own, on the same key. The trace below is one crawl through all three.

authentication $

export SPIDER_API_KEY="<YOUR_API_KEY>"

Get a key →

api.spider.cloud waiting 200 · 30 pages
The API
Step one of an example trace, captured 2026-09-09. The API receives POST /crawl for doc.rust-lang.org/book with a limit of 30 and streams back 30 pages of markdown, 319 kilobytes in all, with status 200. Every page came back 200. demo-01 is an illustrative correlation id, not an API field.
POST /crawl · limit 30 30 pages · markdown · 319 KB
Fetch a page, crawl a site or search the web. Pages stream back as they land.
See the API →
browser.spider.cloud warm open done
The browser
Step two. The browser session was already warm. It opens a session for the API, loads each of the 30 pages from doc.rust-lang.org/book through a residential exit in Germany, and hands the markdown back as each page finishes. demo-01 is an illustrative correlation id, not an API field.
doc.rust-lang.org/book DE · RES
# The Rust Programming Language
The first step is to install Rust. We’ll download Rust through `rustup`, a
command line tool for managing Rust versions and associated tools. You’ll need
an internet connection for the download.
Note: If you prefer not to use `rustup` for some reason, please see the
[Other Rust Installation Methods page](https://forge.rust-lang.org/infra/other-installation-methods.html) for more options.
Our own browser engine opens in milliseconds, with the full native profile and the premium network on.
See the browser →
proxy.spider.cloud waiting DE · RES
The network
Step three. The network routes the browser's requests through a residential exit in Germany, rotating the exit per request, and returns the 30 pages. The pool is 215M+ exits in 199 countries. demo-01 is an illustrative correlation id, not an API field.
215M+ exits · 199 countries
doc.rust-lang.org/book exit · DE · RES
215M+ exits, 199 countries, one per request.
See proxy locations →

Use the whole web like a dataset.

Spider is a research lab for web data quality. Most websites have no API, so we ship one. It crawls, scrapes and searches the web and returns clean markdown or JSON for your RAG pipeline or agent.

render request: "browser"
Renders like a user
Engraving of an eighteenth century printing shop, the pressman at the press with a sheet on the tympan. From the Encyclopedie of Diderot and d'Alembert.
Scripts run, lazy images load, and infinite scroll reaches the end. You get the page a person sees, not the one the server sends.
See the browser →
checkpoint stealth: true
Loads the protected page
Engraving of the Porta Maggiore in Rome, two arches under an inscribed attic with the road passing through. From Piranesi's Vedute di Roma.
Spider is a real browser engine, so a protected page loads for it the way it loads for a person, geo checks included. What comes back is the page, with a 200.
See the scrape endpoint →
extract return_format: "markdown"
Extracts the content
Woodcut of a papermaker lifting a fresh sheet from the vat on a mould, a boy carrying finished sheets beside him. From Jost Amman's Book of Trades.
Navigation, ads and footers drop away. What is left is markdown or JSON that reads the way the page was written.
See the formats →
stream application/jsonl
Streams in order
Engraving of the Marly machine in elevation, a water wheel driving a row of pumps on pilings. From the Encyclopedie of Diderot and d'Alembert.
Each page in a crawl arrives as one JSONL line the moment it is done, in order. Nothing waits for the last URL.
See streaming →

Pay for usage, or for concurrency.

pay as you go

no minimum

$1 per GB

Plus $0.0001 per CPU minute.

Bandwidth: $1 per GB
CPU time: $0.0001 per minute
Blocked, timed out, failed: $0

Start free→

unlimited

flat rate

$40 a month

From 2 in flight. See pricing

Requests: unlimited
Starting concurrency: 2 requests in flight
Per concurrent request: $20 down to $13

See unlimited plans→

Make your first request.

Three steps, the same ones the dashboard shows after you sign up. Pick your client and paste.

quickstart Authorization: Bearer $SPIDER_API_KEY

One key, one request

Every client below sends the same header, so the key is the only thing you type. Connect an agent or call the API directly; the page comes back the same way.

Quickstart in the docs →

Engraving of two hands cutting a quill with a penknife. From the Ecritures plates of the Encyclopedie of Diderot and d'Alembert.

01 Get a key

Create API key

export SPIDER_API_KEY="sk-..."

SDKs, the MCP server and every command read it.

02 Connect Claude Code

claude mcp add spider --transport http \
  https://mcp.spider.cloud/mcp \
  -H "Authorization: Bearer $SPIDER_API_KEY"

CLAUDE.md

For web pages use spider-agent. Read https://spider.cloud/agent-skill/spider-agent/SKILL.md first.

Or the CLI. It picks HTTP or a browser per page, stops at your budget, and returns 187 bytes for named fields where asking for everything returns 9021.

Client configs in the docs →

02 Connect Codex

Read and follow https://spider.cloud/agent-skill/SKILL.md

Paste into a Codex session. The skill covers the whole API.

AGENTS.md

For web pages use spider-agent. Read https://spider.cloud/agent-skill/spider-agent/SKILL.md first.

Or the CLI. It picks HTTP or a browser per page, stops at your budget, and returns 187 bytes for named fields where asking for everything returns 9021.

Every command in the docs →

02 Install the CLI and run it

curl -fsSL https://spider.cloud/install/spider-agent.sh | sh

Downloads the release build, checks its SHA256, and puts it in ~/.local/bin. With a Rust toolchain, cargo install spider-agent-cli works too.

spider-agent scrape https://example.com --goal markdown --json

Picks HTTP or a browser per page, escalates on a refusal, stops at the budget you set.

Every command in the docs →

02 Connect your MCP client

Endpoint:  https://mcp.spider.cloud/mcp
Transport: Streamable HTTP
Auth:      Authorization: Bearer sk-...

Works in Claude Desktop, Cursor, Windsurf, or any MCP client.

Client configs in the docs →

02 Make the first request

curl -X POST https://api.spider.cloud/scrape \
  -H "Authorization: Bearer $SPIDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com", "return_format": "markdown"}'

Full example in the docs →

02 Install the SDK and run it

pip install spider_client

from spider import Spider
app = Spider()  # reads SPIDER_API_KEY
page = app.scrape_url("https://example.com")

Full example in the docs →

02 Install the SDK and run it

npm install @spider-cloud/spider-client

import { Spider } from "@spider-cloud/spider-client";
const app = new Spider(); // reads SPIDER_API_KEY
const page = await app.scrapeUrl("https://example.com");

Full example in the docs →

03 Read the data

example output example data after a scrape

[{
  "url": "https://example.com",
  "status": 200,
  "content": "# Example Domain\n\n..."
}]

The connection comes first. Data like this arrives once the client makes a request.

Markdown or JSON, streamed. Drop it into your RAG store, agent context or database.

Start with one URL.

Create a key. Make your first request.

Get started free

You are visitor 01402503 — GeezerVison does not store the pages you look up.
Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required.

GeezerVison

Infrastructure for reading the public web.

Scraping API, browser, proxy. One key.

The API

The browser

The network

Use the whole web like a dataset.

Renders like a user

Loads the protected page

Extracts the content

Streams in order

Pay for usage, or for concurrency.

pay as you go

unlimited

Make your first request.

One key, one request

01 Get a key

02 Connect Claude Code

02 Connect Codex

02 Install the CLI and run it

02 Connect your MCP client

02 Make the first request

02 Install the SDK and run it

02 Install the SDK and run it

03 Read the data

Start with one URL.