Brave Search API now accepts agentic payments via the x402 protocol
brave.com
brave.com
I do want it to work, so just curious does anyone have any positive anecdotes? Can this naturally get picked up by crawlers?
No one is willingly going to tack on a variable cost to their business revenue model. At best, the crawlers/users will simply skip your hosts, or operate in places where gleaning data is still perfectly legal. =3
It also means that if this starts getting any traction, then we could expect to start a nice bump in prepaid debit card adoption.
- The per-API-call return probably needs to be fairly high with obvious value to justify paying for something. Downloading a .tar.gz of the site, rather than a crawler.
- The world seems to be reforming political blocks, so it'll likely be a cross-bloc transaction where the two sides don't have a fiat transfer system available for use.
- Probably something that is politically sensitive in the customer's bloc.
So my guess would be that the first use case will be some illegal service that could have been implemented with an AWS Lambda, probably hosted somewhere questionable (Russia a candidate). Probably branching out into politically sensitive topics.
> People joking that the only people putting money into x402 are the people testing to see if x402 works.
It'll be that way until suddenly it works for someone, like how bitcoin was a joke up until some tipping point where the financiers started moving in. It'll take a big kick initially to convince people figuring out crypto is worth it just for the sake of making a web trade.
99% of the big companies provide a standardized known way of paying stuff everyone is used to use.
Its a credit card.
Whatever happens somehow in the background, no one cares.
EDIT: And that's without mentioning the value of machine-readable payment requests, which is what x402 actually brings.
Sure it’s a per-vendor setup, so you might need to do this for each 3rd party that requires payment, or use something like openrouter with a pool of funds.
Probably simpler than setting up and using x402.
I'm even less inclined to put tens of dollars into something my agent thinks might be interesting, but I wouldn't be opposed to putting a dollar in a "browsing budget" for the day.
I mean i pay a drink for a euro with a credit card. Whats the big business behind 'pay per visit'?
Eeh, have you find any crawler software that actually supports that protocol? I've seen lots of websites implementing support for it, but do people really implement support for a protocol before they see user-agents who could technically use it? If you're seen those user-agents, which ones are they?
The problem was never the lack of a way to pay you, they just don't care if the transaction is fair.
Admittedly, the price that most will be willing to pay is indeed capped by the cost of the alternatives (bypassing anti-scraping measures via proxies, etc). Still, that means that there's a nonzero rational price.
$5 for 1000 searches. A lot of basic ChatGPT sessions can easily do 30 web searches , which would already cost $0.15 .
This brings up an interesting concern: providing search obviously isn't free. It has been financed through ads thus far. What happens when the majority of searches are done by agents?
Exa is $4 for 1000 searches.
If you want Google results, SerpApi gives 5000 searches for $75. They are not pay-as-you-go, and more searches give you better pricing. Probably if you're setting this up for your whole company you can get a pretty good deal.
Also SerpApi results just as good as Google? Sometimes Google still beats Kagi (maybe I’m search bubbled despite light anonymity measures).
Oh wow that's quite expensive, when I tried to create a similar API to resell google search results I was able to sell 15k requests with $34(and this was the lowest plan with the most expensive price for 1k), with a profit margin of about 70-80%, that was about 2 years ago tho, so maybe now it's much harder, but the price still seems insane. Unfortunately didn't got much traction and I stopped working on it after a few months, since it required a lot of maintenance.
Yeah, that’s where the extra margins go usually - traction (marketing) and maintenance.
What about this gave you the impression it was an unsustainable price?
The price was very sustainable if I would've been able to find enough clients, but I guess the other replies are right that most of the margins and the premium the clients are paying for similar services is going towards marketing/brand awareness.
Would love to hear their thoughts and visions of what the web could be at the time.
The idea of paying for usage is actually very appealing.
People do this intuitively with payments (negative). They'll buy a timeshare or a boat (expensive, "lump" cost) when it would be much cheaper on net to stay at resorts or rent a boat and pay by usage.
They'll pay for subscriptions (automatic payments) rather than pay per experience. Many people will pay $20/mo for Netflix, watch it twice a month, but would never rent movies for $3 on Amazon.
They'll pay a big premium for "unlimited" data plans when this actually makes little sense for anyone vs. paying by usage.
There’s no reason a big host can’t be 100% free, entirely relying on user payments for requests to pay for electricity and new servers (where each request is almost free), a few engineers and security guards to physically install, maintain, and monitor the servers (funded by excess or donations).
Instead the usage could be billed (lumped) at the end of the month and hidden in the moment so there's no real micro-decision or micro-pain response there.
One could easily imagine that, or that your AI provider, being ChatGPT, Anthropic or Cursor or even neo clouds where you run your own compute on, provide a weekly or monthly aggregated payment not just for computes but for paid traffic. They have the balance sheet (sigh) to front it.
You'd pay $20 for compute and $8.7896813 for traffic or you'd have a "budget" of $10 of traffic included in your $30 bundle.
I think customers would accept that if explain that otherwise their queries wouldn't be as nice and also that the providers of content would get a share.
If there is a message-to-the-medium of LLMs and the agentic internet era, it's commodification. More digital services are going to converge on metered billing, with incremental usage being extremely cheap - like the lightswitch, cheap enough to be implemented without the explicit payment-auth step on each minor decision.
Software and users will likely start setting some sorts of policy limits and alerting systems for different categories of spending.
It's definitely gotten better in the last year or two and the AI results at the top seem to be just as good as the Google ones for most topics.
It's also easy enough to switch a search to another provider using "!". Just add "!g" to a search and it will go to google, "!gi" will search google images, there are a TON of those bang commands too.
https://support.brave.app/hc/en-us/articles/4410152384781-Wh...
That is at least 10x too expensive.
If I'm paying for internet already, can I just smear that out and pay for whatever website I'm looking at at the time?
Granted this is for micropayments and each search will most likely cost pennies but it still means loading up some real money into a digital wallet which an agent will blindly (and most likely very inefficiently) spend at its own discretion.
Time will tell, I guess.
You wouldn’t necessarily have subscriptions to all these tools, or one mega car shopping one. This lets your agent build a custom task for you and buy information à la cart.
The x402 landscape right now is “we have an API, let’s throw a price on it”
Mind you Visa could have implemented this when everyone first asked back in 1999, but hey here we are. Rather than micro payments instead we let ads do their best to ruin western civilization.
With a blockchain, no need for all of that. And you can only pay for few cents, no need to top-up a full 5$ min for each services. If you have dozens of services that you need to reach, the benefits start to be interesting.
USDC is much simpler.
Does accepting payment in crypto exempt a merchant from local customer-service requirements?
> You can sue me through legal proceedings but if I win, you pay my legal fees
That's simply not true in at least two jurisdictions I touch, which would have venue for my consumer purchases. (If you never visit the United States or EU, on the other hand, this probably wouldn't practically affect you.)
ToS don’t overrule law.
> lets face it, you are not going to sue for $0.0005 anyway
Very true. Though if I felt cheated by someone, regardless of the cost, I’d absolutely be open to starting a regulatory shitstorm. (Not saying you would cause that among anyone paying you.)
Unless I'm confused, it seems like for a US taxpayer, this mechanism would require that their federal income tax return include, either on an 8949 or an attached statement, an itemized list of every single transaction they made for every single search over an entire year.
I looked through the instructions on the forms, and they say you may round to the nearest dollar. In theory, I think you'd need to compute capital gains (sell price - purchase price) at each transfer, but I expect in practice they'd almost all round to zero since USDC is supposed to be equal to the value of the US dollar.
for 99% of people the first thing they need to do is to understand x402, Purl Wallet, Stripe-developed CLI wallet and base-native USDC.
Funny how crypto tries to snake itself back into our ecosystem again and is still as shitty as it was bfore.
I can't vouch for the quality of the project.
Monero has slow confirmation times. It requires a 0-conf to be able to do micropayments. Most wallets also require a full 10 confirmations before you can use your balance.
So the missing piece for me isn't 402 itself, it's a budget envelope around it: a per-task cap, client-side dedup/caching of near-identical queries, and a receipt ledger I can audit after the run. Without that, making payment frictionless mostly makes runaway spend frictionless too. Does Brave's proxy do any dedup or idempotency for repeated identical queries, or is that entirely left to the harness?
Why are all of these comment bots trained to keep coming back to something?[1][2]