Agent services, priced per call.
Ten HTTP endpoints your agent can call and pay for in the same request. Payment is USDC over the x402 protocol on Base (or MPP), so there's no account, no API key and nothing to invoice. Every call runs on Agent Harbor's own hardware.
- 10live endpoints
- $0.003–$0.05per call
- Basemainnet, USDC
- 20–120 sresponse targets
How payment works.
Your agent pays inside the HTTP exchange itself. Any x402 client library handles the steps below for you.
- Call the endpoint without paying. You get
402 Payment Required, with the x402 v2 terms base64-encoded in thePAYMENT-REQUIREDheader. - Sign a USDC payment for the quoted amount. Scheme
exact, on Base, to the address in the terms. - Retry with the payload in
PAYMENT-SIGNATURE. We produce the result first, then settle. Nothing is charged unless the call is served.
Endpoints priced $0.01 and up also accept the Machine Payments Protocol: the 402 carries a WWW-Authenticate: Payment challenge, paid in Tempo USDC and settled through Stripe.
protocol x402 v2, scheme exactnetwork Base mainnet (eip155:8453)asset USDCpay to 0x840C829530DF7B4CCF700Fa3A0060cF2C80b014ebase url https://x402.agentharbor.agency also MPP on routes from $0.01 (Tempo USDC) free GET /catalog GET /health GET /openapi.json GET /SKILL.md GET /.well-known/x402
Every endpoint.
Prices are per call in USDC. Response targets are the longest a call should take; most return sooner. The live /catalog is the source of truth if anything here differs.
Company and web data
Who's behind a domain and what their site runs on.
-
/company$0.005 USDC
Compact company card: business name, a published generic role contact address, and up to 10 detected technology names. Includes the free content_safety field.
- Inputs
domain- Method
- GET, POST
- Pay with
- x402
- Target
- 45 seconds
- Example
GET /company?domain=example.com
-
/scrape$0.01 USDC
Business scrape: business name with its source, public contact signals, and the platform stack with evidence for each detection. Unsupported claims are struck with a reason. Includes content_safety. Full docs
- Inputs
domain- Method
- GET, POST
- Pay with
- x402 or MPP
- Target
- 45 seconds
- Example
GET /scrape?domain=example.com
-
/enrich$0.05 USDC
Everything in /scrape plus firmographics and a tech-stack summary, written by a model running on our hardware and checked against the evidence. Includes content_safety. Full docs
- Inputs
domain- Method
- GET, POST
- Pay with
- x402 or MPP
- Target
- 120 seconds
- Example
GET /enrich?domain=example.com
-
/domain-verify$0.005 USDC
DNS, RDAP registration, TLS and website checks for a domain, with risk flags.
- Inputs
domain- Method
- GET
- Pay with
- x402
- Target
- 30 seconds
- Example
GET /domain-verify?domain=example.com
Site audits
What a website tells search engines, ad platforms and AI crawlers.
-
/audit$0.05 USDC
Home-page audit of analytics tags, consent, structured data, AI-crawler access and SEO, returned as a ranked fix list.
- Inputs
domain- Method
- GET
- Pay with
- x402 or MPP
- Target
- 75 seconds
- Example
GET /audit?domain=example.com
-
/ai-crawlers$0.01 USDC
Which AI crawlers, such as GPTBot, ClaudeBot, PerplexityBot and Google-Extended, may fetch the home page under robots.txt, plus sitemap, llms.txt and noindex checks.
- Inputs
domain- Method
- GET
- Pay with
- x402 or MPP
- Target
- 40 seconds
- Example
GET /ai-crawlers?domain=example.com
Public records
Government data, cleaned up for machines.
-
/sec-company$0.01 USDC
SEC EDGAR profile, XBRL financials and recent filings for U.S.-listed companies, looked up by ticker or CIK.
- Inputs
ticker,cik,filings,forms- Method
- GET
- Pay with
- x402 or MPP
- Target
- 30 seconds
- Example
GET /sec-company?ticker=AAPL
-
/recalls$0.005 USDC
Recall matches from CPSC, NHTSA and FDA, by keyword or by vehicle make, model and year.
- Inputs
category,query,make,model,year,limit- Method
- GET
- Pay with
- x402
- Target
- 30 seconds
Developer tools
Small checks a coding agent needs mid-task.
-
/package-check$0.003 USDC
Known vulnerabilities from OSV, license and deprecation status for an open-source package version.
- Inputs
ecosystem,name,version- Method
- GET
- Pay with
- x402
- Target
- 20 seconds
-
/pdf-extract$0.01 USDC
Text of a public PDF as markdown, tables included, up to 30 pages per call.
- Inputs
url,pages- Method
- GET
- Pay with
- x402 or MPP
- Target
- 60 seconds
Other ways to find and buy.
The gateway is the most direct route, but it isn't the only one.
-
Apify Store
PublishedOur website tracking, consent, schema and AI-crawler audit runs as a pay-per-event Actor, for Apify users and the agents they run.
-
x402 Bazaar
ListedCoinbase's directory of paid x402 services. Agents searching there are routed to our gateway.
-
PayAPI Market
Listed/scrape, /company, /enrich, /audit and /ai-crawlers, settlement-verified and searchable through the market's MCP endpoint.
-
MPPScan
ListedEvery route that accepts the Machine Payments Protocol.
-
Discovery files
PublishedMachine-readable descriptions of every service: the A2A agent card, agent index, llms.txt, the gateway's x402 manifest and OpenAPI spec.
Questions from agent builders.
Do I need an account or an API key?
No. Payment is the authorization. Any caller with a funded wallet and an x402 client can use every paid endpoint.
What does my agent need to pay?
A wallet holding USDC on Base mainnet and an x402 v2 client that can sign an exact payment. Agents set up for the Machine Payments Protocol can pay the $0.01-and-up routes with Tempo USDC instead. There's no minimum balance and nothing to top up with us.
Am I charged if a call fails?
No. We produce the result before the payment settles, and nothing is charged unless the call is served.
Where does my input go?
To our machines in Bremerton, Washington. Models run on our own hardware, so what you send isn't passed to a third-party AI provider. Public-records endpoints read the public source they name, such as SEC EDGAR.
How do I know a result is trustworthy?
Our scrape and enrichment responses carry a validation block: each platform claim is scored against the page evidence behind it, unsupported claims are struck with a reason, and the response names the model, time and gate version that checked it.
What about prompt injection in scraped pages?
/scrape, /enrich and /company include a free content_safety field that flags injection patterns in the site text we return, with the JSON path, the rule that matched and an excerpt.
Can a person use these without writing an agent?
Yes. Any x402 client works from a script, and the site audit is also available on the Apify Store with no crypto wallet needed.
I need data you don't sell yet.
Email anchor@agentharbor.agency with what your agent needs and what it would pay per call. New endpoints come from requests like that.
Building something on top of us?
We'd like to hear about it, especially if you need higher volume or a new endpoint.