Skip to content

Cheaper web scraping for AI agents.

Stop paying your model to scrape. Clean web data, at scale, for a fraction of the cost.

Powering open models with bandwidth.

Terabyte-scale web data for open source models, over a residential network, for a fraction of the cost.

❯_runner-ai · claude code
 Scrape example.com as seen from Germany and clean it

 I'll use the Runner AI scrape skill to fetch example.com from Germany and clean it.

 Skill(runner-ai:scrape)
   Successfully loaded skill

 exit acquired · DE · 38 ms
 clean content · 448 B · status 200

 Done. Routed through a residential node in Germany (DE), noise tags stripped out.
Result: a minimal placeholder page (448 B · 0 credits used).
- Title: Example Domain
- Link: Learn more → iana.org/domains/example

 Worked for 21s · no API key
Built forClaudeChatGPTGemini
Public Data for Your Open-Source Model Today
  • Kimi
  • GLM
  • DeepSeek
  • Qwen
  • Llama
  • Mistral

Built for clean data, priced like infrastructure

Geolocated to any region

Only Runner

Pin any request to a country and pull data from a real residential IP there. Native, end-to-end geolocated coverage across the network, not a proxy bolt-on after the fact.

Prepaid pools, never expiring

You buy bandwidth in gigabytes as a prepaid pool, not metered credits that burn down. Pools never expire, top up anytime, and route across a network of real residential IPs.

Priced to undercut the incumbents

A residential network with no server farms to mark up and no middleman margin. The structural result: per-GB rates that undercut list-price incumbents several times over.

Data delivery at scale, without the incumbent markup

Coverage in every region you sell

Only Runner

Pull data from real residential IPs in any country your business operates in. Native geolocated coverage across the network means one vendor for every market, not a patchwork of regional proxy contracts.

No contracts, no monthly commit

Buy bandwidth as a prepaid pool that never expires. No annual commit, no seat licenses, no burn-down credits. Scale spend up or down with the workload instead of a procurement cycle.

A structurally lower cost basis

A residential network with no server farms to mark up and no middleman margin. One million HTML pages for the cost of a lunch, and multimedia datasets no centralized vendor will quote at any price.

A request is a few lines and an API key

Add credit once, then pull data. One API key, no dashboard to provision, no per-call setup. The HTTP API is live today; the typed clients below are what we are building next.

# Live today — this is the shipped endpoint.
curl -X POST https://api.runner.dev/v1/scrape \
  -H "Authorization: Bearer $RUNNER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/pricing",
    "country": "US",
    "formats": ["markdown", "links"]
  }'

# => { "success": true, "data": { "markdown": "...", "links": [...] },
#      "credits": { "charged": 2.4, "balance": 97.6 } }
Agent-native

Hand your agent the open web

Runner speaks MCP, so any agent picks up scraping, geolocation, and extraction as first-class tools. No glue code, no scraper to babysit — one credential and the web is a callable surface.

  1. 01

    Register the toolset. One line wires the MCP server into Claude Code, Cursor, and the rest.

  2. 02

    Let the agent bootstrap itself. Point an autonomous agent at the skill and it provisions its own key, then starts calling.

runner ~ agent shell
01MCP
runner> npx -y runner-cli@latest init --all --browser
Read the setup guide
02cURL
runner> curl -s https://runner.dev/skill
Inspect the skill

Four categories. One transparent rate each, discounted by volume.

Pick the level of service you need. Every category starts at 15% below list price with no commitment, and the discount deepens the more you buy. No credits, no multipliers.

Web Scrape

$0.71/1K pages

Full page fetch through a residential IP. Returns clean HTML, plain text, or Markdown with tag stripping included. No JavaScript rendering.

Best for: Static pages, blogs, articles, documentation, product listings.

Dynamic Scrape

$2.83/1K pages

A headless browser renders the page through a residential IP. Handles SPAs, infinite scroll, and lazy-loaded content. Returns fully rendered HTML, text, or Markdown.

Best for: Modern web apps, social media, dynamic dashboards, JS-heavy sites.

AI Extract

$3.53/1K pages

Scrape plus LLM-powered extraction. Returns structured JSON matching your schema — tables, product data, pricing, reviews, contact info, anything you define.

Best for: AI training data, price monitoring, lead generation, dataset building.

Multimedia

$0.50/GB

Direct download of binary content through a residential IP. No rendering or parsing — pure bandwidth for large files, billed flat with no discount ladder.

Best for: AI training datasets, media monitoring, PDF archives, research.

One API behind a lot of different jobs

The same clean, geolocated data feeds very different pipelines. A few of the things teams point Runner at.

AI training data

Pull large, geolocated datasets — HTML, text, and multimedia — to feed model training and evaluation. Priced so a full dataset does not blow the budget.

Price monitoring

Track pricing and availability across markets and regions on a schedule, with structured extraction into your own schema.

Lead generation

Extract contact and company data from directories and listings into clean, structured JSON ready for your CRM.

Dataset building

Assemble structured datasets from product pages, reviews, and listings — tables and fields you define, returned as JSON.

Agent tooling

Give autonomous agents a web-data tool via the MCP server: scraping, geolocation, and extraction as first-class capabilities.

One vendor behind a lot of programs

The same clean, geolocated data feeds very different programs across an organization. A few of the ways teams run Runner at scale.

Training-dataset acquisition

Acquire terabyte-scale, geolocated datasets — HTML, text, and multimedia — for model training and evaluation, at a cost basis no centralized vendor can match.

Market intelligence

Monitor pricing, availability, and positioning across every market you operate in, on a schedule, with structured extraction into your own systems.

Data enrichment

Enrich CRM and sales records at scale with company and contact data pulled from directories and listings into clean, structured JSON.

Proprietary datasets

Build and refresh proprietary structured datasets from product pages, reviews, and listings — the fields you define, delivered as JSON.

Internal agent infrastructure

Give internal agents and automations a shared web-data capability via the MCP server: scraping, geolocation, and extraction as governed, first-class tools.

Buy a Data Pool. The discount deepens with volume.

Every tier undercuts list price — from 15% off with no commitment up to 50% off at Infrastructure. You buy bandwidth, not credits. Pools never expire and you can top up anytime.

Starter

$29/ pool

25% off included

Effective rate
$0.62 / $2.50 / $3.11
Per 1K pages
Scrape / Render / Extract per 1K
Throughput
25 concurrent
Enter
Most popular

Pro

$99/ pool

35% off included

Effective rate
$0.54 / $2.17 / $2.70
Per 1K pages
Scrape / Render / Extract per 1K
Throughput
50 concurrent
Enter

Scale

$499/ pool

42% off included

Effective rate
$0.48 / $1.93 / $2.41
Per 1K pages
Scrape / Render / Extract per 1K
Throughput
200 concurrent
Enter

Infrastructure

$1,999/ pool

50% off (max) included

Effective rate
$0.42 / $1.67 / $2.08
Per 1K pages
Scrape / Render / Extract per 1K
Throughput
Unlimited concurrency
Enter

15% below list price. No commitment. $5 minimum to start.

Add credit and start immediately. Failed requests are never charged.

ServicePer GBPer 1K pages
Web Scrape$0.71$0.71 / 1K pages
Dynamic Scrape$2.83$2.83 / 1K pages
AI Extract$3.53$3.53 / 1K pages
Multimedia$0.50

Scrape / Render / Extract rates per 1,000 pages, discount vs. list price. Multimedia is billed flat at $0.50/GB, no ladder. Failed requests are never charged.

What the sticker price hides

Not subsidized burn. A residential network has no server farms to mark up and no middleman margin, so almost all of your payment reaches the infrastructure delivering the data.

WorkloadLeading providersCentralized providersRunner
AI Extract, real usage$7.76 / 1Kvaries by tier$3.53 / 1K
Extract, max discount$7.76 / 1Kvaries by tier$2.08 / 1K
Wholesale IP, entry tier$8-15 / GB$8-15 / GB$0.99 / GB
Wholesale IP, top tiercustom quotecustom quote$0.40 / GB

A $83/mo credit-based plan resolves to $7.76 per 1,000 real extractions after failure overhead — 2.9x Runner PAYG, 3.7x Runner at max discount. No credit system, no failure tax.

15% to 50% cheaper. Structurally, not a loss leader.

Web scraping — cost per 1,000 pages

Credit-based platforms$0.83list price
Runner — PAYG (no commitment)$0.7115% off
Runner — Infrastructure pool$0.4250% off (max)

JS rendering — cost per 1,000 pages

Credit-based platforms$3.33list price
Runner — PAYG (no commitment)$2.8315% off
Runner — Infrastructure pool$1.6750% off (max)

Not subsidized burn. A residential network has no server farms to mark up and no middleman margin, so almost all of your payment reaches the infrastructure delivering the data.

An MCP server and a plugin for every workflow

Wire Runner into your agent with the MCP server, or drop the browser plugin into your stack. Same API key, same pools, zero config.

MCP server

Give any MCP-aware agent scraping, geolocation, and extraction tools. One endpoint, your API key as the credential.

Join the MCP waitlist

Browser plugin

Scrape and extract straight from the page you are on. The plugin routes through the same residential network and prepaid pools.

Join the plugin waitlist

The answers before you ask

How does pricing work?

You buy bandwidth measured in gigabytes as a prepaid pool, not metered credits that burn down. Bigger pool, lower per-GB rate. Pools never expire, and you can top up anytime. Pay-as-you-go starts at a $5 minimum.

Do I pay for failed requests?

No. Failed requests are never charged. You pay only for data actually delivered — no credit multipliers, no per-call ceiling, no hidden tiers.

Can I target a specific country or region?

Yes. Pin any request to a country and pull data from a real residential IP there. Geolocated coverage is native to the network, not a proxy bolt-on after the fact.

What formats do I get back?

Clean HTML, plain text, or Markdown for scraping; fully rendered HTML for dynamic pages; and structured JSON matching your own schema for AI extraction. LLM-ready output is the default, not an add-on.

How do I get access?

Access is gated while we onboard early members. Request access to get product updates and your first keys when the network opens — no application, no sales call.

Is there a contract or minimum commit?

No contracts, no minimums, no annual commit. Add credit, pull data, and pay only for what you use. One API key, no lock-in.

How does pricing work at volume?

You buy bandwidth in gigabytes as a prepaid pool — bigger pool, lower per-GB rate, down to a fraction of a cent per gigabyte at scale. Pools never expire, there is no annual commit, and you top up as the workload demands.

Is there a contract or minimum commit?

No. No contracts, no seat licenses, no annual commit. Spend scales up or down with your workload instead of a procurement cycle. Failed requests are never charged.

Can you deliver in every region we operate in?

Yes. Pull data from real residential IPs in any country you sell in. Geolocated coverage is native to the network, so one vendor covers every market — no patchwork of regional proxy contracts.

Can you handle large training-dataset volumes?

Yes. The network is priced for scale — one million HTML pages for the cost of a lunch, and multimedia datasets no centralized vendor will quote at any price. Custom pools and dedicated coverage are scoped on a call.

How do we get started?

Book a call and we will scope your volume, geolocation, and delivery, then set up custom pools for your team. Because the clients are open source, your security and diligence teams can review exactly what runs in your stack.

What formats do we get back?

Clean HTML, plain text, or Markdown for scraping; fully rendered HTML for dynamic pages; and structured JSON matching your own schema for AI extraction. LLM-ready output is the default, not an add-on.

Transparent pricing, no lock-in

No contracts, no minimums

No applications, no account managers, no annual commit. Request access, get a key, and pay only for what you pull.

Pools never expire

You buy bandwidth measured in gigabytes, not credits that burn down. Your balance is yours until you use it.

Pay only for data delivered

Failed requests are not charged. No credit multipliers, no per-call ceiling, no hidden tiers.

Built to stay online

A broad residential network with requests routed across many IPs and regions, engineered for uptime under real-world load.

Request access before the network opens

Access is gated while we onboard early members. Request access for product updates and first keys when the network goes live.

Talk to us about data at scale

Book a call for at-scale delivery, custom pools, and dedicated geolocated coverage. We will reach out to scope your volume.

One API for web data. Get your key.

Request access, add credit, and start pulling clean, LLM-ready data. No contracts, no minimums, no sales call required.

The data you need, at a price no one can match.

A terabyte of web data for the cost of lunch. Book a call and we will scope your volume, geolocation, and delivery.