URL IN. MARKDOWN OUT. BUILT FOR AGENTS.
Our model Iris gives your agent eyes.Tables, icons and the hidden JSON, suddenly visible.
Hacker News
Figma pricing
Our World in DataWhat should your agent see next?
looks.Hidden frontend state delivered as clean JSON
Nav, footer, ads, banners gone before you pay for them
CSS/SVG icons become text your LLM reads
Your agent sees what ordinary markdown APIs miss.Ordinary converters keep clutter, drop the table, and never see what only rendered in the browser.Iris sees the rendered page and returns three things: markdown with tables and icons intact, the hidden JSON, and an appendix with everything else.Nothing lost, nothing in the way.
Every line links back to where it came from, so you can verify what your agent saw.
Hover to debug. Share to cite.
It works out of the box without any configuration—JavaScript rendering, residential proxies, and bot protection are handled automatically for every request.
Even with accurate markdown, a 10k token page makes your agent slower, more expensive, and more likely to lose focus. Pass a query to Iris and get back only the facts your agent needs.
Faster TTFT
Zero noise
const { markdown } = await expand.fetch(
'https://stripe.com/docs/api',
{ query: 'webhook signature verification' }
)Paste one prompt into your coding agent. Expand will configure every supported harness detected on this machine, then verify the setup with an exact citation.
A note from the founder
Your agent gets it wrong, or says the page doesn't say. You open the page. It's right there, in a table the agent never saw. Model or scraper? You can't tell.
We build research agents and lived there for months. Footers dropped, and the company address with them. Pages that never loaded. Things I could see and the agent couldn't. Nobody could check. We tried better heuristics. A rule that works on one site breaks on the next, and every markdown API runs on hardcoded rules. Cognition wasn't the bottleneck. The tooling was.
That's why we built Expand Fetch, the only lossless and verifiable web data API. Iris sees the page the way you do and returns it whole. Every line links back to the exact spot it came from. You click, and there it is.
You ship the agent. Your users get answers they can check.
Tim
Tim Suchanek, founder
Iris removes the trash on a page far more accurately than hardcoded heuristics do. Fewer junk tokens on every request, and that adds up.
Spatial Markdown + JSON
50%
Targeted Query
50%
Raw markdown (others)
37,218 tokens/req
Your blended average
15,818 tokens/req
Monthly savings
$3,210
Annual savings
$38,520
Try the platform at no cost.
$10 in credits to explore everything below.
Stop feeding your agent a broken web.
system.one / YCombinator / Conviction / Pioneer Fund / basecase
Guillermo Rauch / Swyx
And more...
Pricing that scales with you.
No monthly commitments.
Web-scale with white-glove support at enterprise security.
Try the platform at no cost.
$10 in credits to explore everything below.
Pricing that scales with you.
No monthly commitments.
Web-scale with white-glove support at enterprise security.
Expand's concurrency is controlled by tiers.
These tiers are unlocked permanently, for successful spend and time since first successful transaction:
| Tier | Qualification | Concurrency Limit |
|---|---|---|
| 0 | Free Tier | 2 |
| 1 | $10 in successful spend | 5 |
| 2 | $50 in successful spend + 7 full days since first successful transaction | 10 |
| 3 | $200 in successful spend + 14 full days since first successful transaction | 20 |
| 4 | $500 in successful spend + 30 full days since first successful transaction | 50 |
| 5 | $1000 in successful spend + 60 full days since first successful transaction | 100 |
Example: If I have $250 in successful spend:
Need instant concurrency without waiting for the tier to upgrade?
Just $1 per browser a month, up to 100 browsers:
Nr. of browsers
40
Price per month
$40
Perfect for short or long bursts. Cancel anytime, and you keep the extra concurrency through the end of the billing month (prorated monthly).
Contact us for more info.