Hi BlockRun team — I operate AgentWork and would like to propose it as a data-provider integration for research agents using BlockRun.
AgentWork returns public source text with timestamps and SHA-256 hashes. Its web-change endpoint lets a caller retain a baseline and purchase a compact check before deciding whether another model analysis is needed.
Live endpoints (x402 v2, USDC on Base or Algorand; no AgentWork account/API key):
| Endpoint |
Price per successful call |
Result |
| POST /v1/web/read |
$0.005 |
Web text, headings, links and source metadata |
| POST /v1/web/changes |
$0.005 |
Changed/unchanged against caller-supplied baseline; optional line differences |
| POST /v1/pdf/read |
$0.01 |
Embedded PDF text, page numbers, hashes and extraction warnings |
The example takes a saved AgentWork read, verifies its text hash, creates a bounded research prompt with provenance and limitations, and optionally calls the buyer's chosen model using @blockrun/llm. Preview is offline; extraction and model charges are separate. It does not fetch or execute instructions embedded in source material.
Limits: no JavaScript rendering, OCR, logins or paywall bypass. PDFs are limited to 5 MiB/20 pages; web responses to 1 MiB. Source failures are not settled. Customer text and baselines are not retained by this product. Receipts and request counters are retained.
AgentCash discovery resolves all three endpoints with prices and without an API-key requirement. AgentWork settlements have been independently verified using operator-funded tests on both chains. The BlockRun adapter is tested with a mocked SDK; a paid end-to-end BlockRun call has not been performed. We are an early service and are not presenting our internal tests as customer traction.
Would this fit a community example or your data-provider onboarding? We can adapt metadata or interface requirements if needed.
Contact: rookcomps@gmail.com
Hi BlockRun team — I operate AgentWork and would like to propose it as a data-provider integration for research agents using BlockRun.
AgentWork returns public source text with timestamps and SHA-256 hashes. Its web-change endpoint lets a caller retain a baseline and purchase a compact check before deciding whether another model analysis is needed.
Live endpoints (x402 v2, USDC on Base or Algorand; no AgentWork account/API key):
The example takes a saved AgentWork read, verifies its text hash, creates a bounded research prompt with provenance and limitations, and optionally calls the buyer's chosen model using @blockrun/llm. Preview is offline; extraction and model charges are separate. It does not fetch or execute instructions embedded in source material.
Limits: no JavaScript rendering, OCR, logins or paywall bypass. PDFs are limited to 5 MiB/20 pages; web responses to 1 MiB. Source failures are not settled. Customer text and baselines are not retained by this product. Receipts and request counters are retained.
AgentCash discovery resolves all three endpoints with prices and without an API-key requirement. AgentWork settlements have been independently verified using operator-funded tests on both chains. The BlockRun adapter is tested with a mocked SDK; a paid end-to-end BlockRun call has not been performed. We are an early service and are not presenting our internal tests as customer traction.
Would this fit a community example or your data-provider onboarding? We can adapt metadata or interface requirements if needed.
Contact: rookcomps@gmail.com