Crawler API · Web research · Video index
Real-time web & video for AI agents.From real devices, not data centers.
Multi-engine, multi-geo crawling and research: any page, from thousands of phones and laptops in 60+ countries. Plus search inside public video, down to the second. One API, one MCP server.
- Real devices worldwide
- Multi-engine
- Multi-geo
- Open source
- Earn UPT tokens
# Fetch a page from a real device in Germany, as markdownnpm i -g uprockuprock auth login --setup-ai # emails you a code, saves your keyuprock ai crawl https://example.com/pricing \ --country DE --content# Out of credits? uprock ai buy 5# MCP server: https://mcp.uprock.ai/mcp# Header: Authorization: Bearer $UPROCK_API_KEYcrawl_fetch({ "url": "https://example.com/pricing", "country": "DE", "device_type": "desktop"})curl -s https://edge.uprock.com/v1/crawl \ -H "Authorization: Bearer $UPROCK_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "url": "https://example.com/pricing", "country": "DE" }'# → {"status_code": 200, "title": "…", "markdown": "…", "credits_used": 1}# No key yet? Leave out the Authorization line: a few free crawls a day.# Takes 5-20 s: a real browser on a real device loads the page. Nothing is cached.Try it live · no key needed
One request. Many countries. See what changes.
URL to crawl, anything else to research. We ask real devices in several countries at once and show the first three to answer. A crawl opens the page in a full browser and runs its JavaScript, like a person visiting. Nothing comes from a cache, so a crawl takes 5 to 20 seconds.
Try: · ·
Set up your agent
Paste one line into your agent.
Claude Code, Codex, Cursor, OpenClaw or any agent with a terminal. It reads the setup, connects the MCP server and runs a test crawl.
Set up UpRock for me: read https://uprock.ai/agents.md, connect the UpRock MCP server, and run one test crawl. Tell me what it costs before you spend anything.
Or read the files yourself
- /llms.txtIndex of every page
- /llms-full.txtEverything in one file
- /agentsMCP, skill, CLI setup
- /skill/uprock-skill.zipAgent skill
Are you an agent?
Fetch the setup and follow it.
Endpoints, auth, the MCP server, pricing and how to get your human an API key, in one short markdown file.
curl -s https://uprock.ai/agents.mdTry it without a key
Leave out the API key and POST /v1/crawl still answers: a few test crawls a day, so an agent can check it works before anyone pays. Expect 5 to 20 seconds: a real browser on a real device loads the page and runs its JavaScript. Nothing is cached.
Why UpRock
We don't run proxies. We run devices.
A proxy gives you an IP address. A Rocket, our app running on a community member's device, gives you a whole computer: a real browser on a real device, owned by someone in our community who's paid per task. No data centers, no proxy pools, no fleet standing by. That's why the price is low. Use us on our own, or next to Firecrawl, Exa and the rest for the pages they can't get.
01
Real devices
Thousands of real devices worldwide. Sites see ordinary visitors, so you get the page a person would get.
02
Multi-geo
Pick the country per request. See local prices, local search results and region-locked content.
03
Multi-engine
Web research fans out across several search engines at once and merges the results.
04
Best price, no overhead
No free tier, because bots farm them. Just a few test crawls a day with no key, and one low price for everyone.
Pricing
One price. Pay with UPT and pay 20% less.
No free tier, no plans, no sales call. Every tool spends prepaid credits. Buy credits by card or USDC, or swap UPT for credits and pay 20% less. Running a Rocket is the only way to earn them free.
- CrawlA rendered page from a real device, as markdown$0.50/ 1K pages$0.40in UPT
- Crawl on a mobile deviceThe same, from a real phone; fewer of these on the network$0.75/ 1K pages$0.60in UPT
- Web researchOne multi-engine search from the country you choose$1/ 1K queries$0.80in UPT
- Video searchOne search over the shared video index$1/ 1K queries$0.80in UPT
- Video indexingPrivate betaAdd public video to the shared indexContact us for early access →$0.02/ min$0.016in UPT
- VerifyCheck a site works from real devices in several regions: status, screenshot and load speed from each$1.50/ 1K checks$1.20in UPT
Why we can price it like this
The power of a real-device community network.
What real people see
Every page loads on a real phone or computer in the country you choose. Your agent sees exactly what a person there sees, not the blocked or watered-down page a data center bot gets served.
A community, not a server farm
People around the world share their real devices and help each other unblock the web. More devices means more capacity, so the price per page keeps falling as the network grows.
No free tier to pay for
We don’t spend money fighting bots that farm free accounts. That saving goes straight into the price you pay.
Our mission is an open web. UpRock is a DePIN, a decentralized physical infrastructure network: devices people own, doing the crawling, the compute and now the video. We're building a future where every agent can reach the web freely, in real time, at a price anyone can afford.
$100 buys
200,000
pages at $0.50 per 1,000
Sign in to buy creditsCredits are what every tool spends. Any amount from $5 to $1,000, by card or USDC. Credits never expire. Failed requests aren’t billed.
| Firecrawl | 166,666 | You get 20% more here |
|---|---|---|
| Exa | 100,000 | You get 100% more here |
| Tavily | 62,500 | You get 220% more here |
| UpRock, paid in UPT | 250,000 | 20% less than card |
| Swap UPT for credits in your account to get this price: Pay with UPT · How to buy UPT | ||
Others shown at their lowest published price, checked 2026-10-21. Some need a monthly plan.
Your account is a wallet of credits.
Every tool spends from it at the prices above, however you fill it.
01
Buy once
Card or USDC, from $5. One price in every country.
02Soon
Top up automatically
Monthly, or whenever the balance runs low, so an agent never runs dry. Same price per credit.
03
Swap UPT: 20% off
Buy or earn UPT, send it to your account and swap it for credits. Every credit costs 20% less than by card.
How to buy UPT →04
Run a Rocket
1,000 credits over your first 10 days with the npm daemon, plus UPT for the work it does.
Contributors: 20% of what you spend indexing public video comes back as credits for video search.
Compare
Our prices next to Firecrawl, Exa, Tavily and the rest →
Crawl, web research, video search and indexing, each linked to their pricing page.
Agent calculator
Building an agent? See the cost per user per month →
Pick how many users you have and how active they are.
Need more than the API?
Advanced jobs on the network
Anything a person could do in a browser, as a custom job on real devices: sites that block datacenter crawlers, multi-step flows with downloads and screenshots, watching pages from chosen countries on a schedule. Every job is scoped and quoted. Rough range: setup $500–$5,000 depending on sites and steps, and 2–20 credits per page instead of 1.
Tell us about the jobEarn
Run a Rocket. Get the best price.
Your machine earns UPT. Swap it for credits on uprock.ai and your agents pay 20% less for crawling, web research and video search. An agent can run a Rocket itself and pay for its own web access. UpRock is a DePIN: every Rocket is a real device owned by a real person, and together they crawl, compute and now index video for the open web.
01 · With a window
Or use the UpRock Mining App.
Same network, same account, with a desktop app and an earnings dashboard instead of a terminal.
Both ways, and what a Rocket does →02 · Have your agent do it
Three commands.
Installs the node from npm, signs in to your account and runs it in the background. It checks status and earnings from the same CLI.
npm i -g uprock
uprock auth login
uprock daemon installGive your agent eyes
All the world's video, searchable.
Your agents need to watch video.
Video is the bridge between the real world and the internet. What people buy, how they react, what happened today: it's filmed and posted long before anyone writes it down. Text tells part of the story; video carries the most context, and context is everything for an agent.
We believe in video that much. So we're building the open video index: one public index across every platform, built by a community network and searchable by any agent, down to the second.
- Big platforms
- Index their own video. Keep it.
- Everyone else
- Every agent pays to process the same clip again.
- UpRock
- Indexed once, searchable by every agent. Shared and open.
# MCP tool on https://mcp.uprock.ai/$UPROCK_API_KEY/mcpvideo_search({ "query": "rooftop pool hotel lisbon" })# → moments: video URL, start/end time, transcriptSearch pricing
- Card or USDC$1/ 1K queries
- In UPT, or after indexing$0.80/ 1K queries
Browse what's indexed for free. Search it by meaning with credits. No storage fees.
1
Search it
Ask for a moment in plain language. Get the video, timestamp and transcript.
2
Feed it
Pay to index a channel, playlist or link. It joins the shared index for everyone. Private beta: contact us for early access.
3
Earn with it
Run a Rocket to earn UPT, then swap it for credits and pay 20% less.
Bring your agent
Claude Code · Codex · Cursor · OpenClaw · Hermes Agent · Venice-powered agents. Anything that speaks MCP or runs a shell command.
Fuel Engine · Open source
When your other crawlers break, send it to the Fuel Engine.
The internet is hostile. Our community swarm is built for it.
The Fuel Engine is what runs under our Crawler API, and it's open source.
01
Route
Each request goes to a real device in the country you asked for.
02
Render
The page loads in a real browser, the same way it would for a person.
03
Recover
If a page is blocked, the job retries on another device or region.
04
Return
You get clean markdown or HTML.
Rate-limited every few minutes? Blocked with 403s? Pages that only render with full JavaScript, or show different content in every country? That's the work other crawlers give up on, and it's the work we do on real devices.
Have a crawl job that keeps breaking? Describe it. We'll tell you if we can do it.
Usually yes. Agents can send the same thing to POST /api/jobs. How the Fuel Engine works.
FAQ
Questions agents ask us.
Is UpRock an alternative to Firecrawl or Exa?
Use it on its own or alongside them. Send UpRock the pages that get blocked, geo-gated or served differently to bots.
Why is there no free tier?
Agents now do most crawling, and they beat CAPTCHAs and open accounts faster than any check can stop them, so free credits get farmed. Every request runs on a paid device, so farming is a real cost. We put what we spent fighting it into one low price instead, and the network itself is cheap to run: devices our community already owns, paid per task, with no data centers or proxy pools to rent. You can still try it: a few keyless test crawls a day, or run a Rocket to earn credits.
Do you use proxies?
No. Requests run on real desktop and mobile devices whose owners opted in and earn UPT for it.
What is the Fuel Engine?
The runtime that runs crawl tasks on those devices: signed, encrypted WebAssembly in a sandbox. The Android engine is open source under Apache-2.0.
Can an agent sign up and pay on its own?
An agent can install the CLI, sign in with a code emailed to you, and run a node to earn UPT. Buying credits by card needs you once.
Does the Video Index copy videos?
No. It records what happens in each moment (transcript, a few small keyframes and an embedding) and links back to the source.
Which countries and devices are supported?
60+ countries, on desktop or mobile devices. The full list is in the MCP tool schema.
Get started
First page in under a minute.
1
Get credits
Buy them by card, USDC or UPT, or run a node to earn UPT.
2
Get your key
Run uprock auth login --setup-ai, or create one in the dashboard and set it as UPROCK_API_KEY.
3
Make a call
Crawl a URL from the CLI, MCP or HTTP. Markdown comes back.
Everything is in one file.