SearchCans

SearchCans provides search and source-extraction APIs for applications, AI agents, and retrieval workflows. Google and Bing search return structured results, while the Reader API converts web pages, PDFs, and Office documents into Markdown or HTML. Both use a shared API key and credit balance.

Retrieve the result and the source separately

A search response identifies relevant URLs; the Reader API then extracts content from a selected source. SearchCans' quick-start example demonstrates that two-step sequence. A result snippet alone is not the same as the article or document behind it, so keep the source URL with extracted text when building a cited answer.

The publisher also lists shopping, image, news, and video search paths, plus file extraction and screenshots. Check the endpoint-specific documentation rather than assuming every option appears in the standard search response. Some pricing-table features are marked coming soon and should not be treated as available production capabilities.

Reader requests can include waiting for JavaScript and proxy-related options. Test extraction against the pages your application uses, including pages with access restrictions or changing layouts. Clean Markdown is a processing format, not a guarantee that the complete source was recovered or that its contents are correct.

Budget for both API calls

The pricing page lists standard Google and Bing search at one credit per request. Reader standard mode uses two credits; shared-pool proxy mode uses four total, with other options costing more. A search followed by reading one page therefore consumes more than a single standard search.

Registration includes 100 free credits without a payment method. Prepaid packs start at $18 for 20,000 credits and do not renew automatically. Credits last six months. The advertised rate starting at $0.56 per thousand applies to larger-pack pricing, not every purchase or every endpoint's effective cost.

The publisher says credits are deducted for successful HTTP 200 responses. Your application still needs to inspect returned data for completeness and relevance; an HTTP success code does not prove an extracted page contains the evidence you wanted.

Size concurrency without assuming throughput

SearchCans calls concurrent request slots Parallel Lanes. Its Standard pack lists two lanes, while eligible larger packs can stack lanes in one account. The service advertises no hourly caps, but the number of simultaneous requests remains bounded by those slots.

Hourly throughput estimates assume a three-second average response time. Actual throughput varies by endpoint and latency. Measure your own workload before treating an estimate as capacity, and handle timeouts and unsuccessful requests. The published 99.9 percent monthly uptime target is described as a service-level objective, not a guarantee. Refund eligibility also depends on requesting within seven calendar days and using less than ten percent of the applicable credits, with fee conditions in the terms.