Skip to content
rivrrivr

Backends — capability, cost & risk

rivr normalizes one schema across backends. Pick one with --provider <name> (or RIVR_PROVIDER); the default is serpapi. This page mirrors the machine-readable profiles from rivr provider list --json — that command is the source of truth, so an agent can choose programmatically.

Pricing/eligibility is indicative (2026) and changes — verify on each provider’s site.

Capabilities

Section titled “Capabilities”
capability serpapi (default) rainforest creators (official) scrape (opt-in) stub
search ✓ ✓ ✓ ✓ ✓ (fake)
item get ✓ ✓ ✓ ✓ ✓ (fake)
offers ✓ ✓ ✓ ✓ ✓ (fake)
reviews ✓ sample ✓ full ✗ ✓ full ✓ (fake)
variations ✓ ✓ ✓ ✗ ✓ (fake)
browse nodes ✗ ✗ ✓ ✗ ✓ (fake)

A capability a backend lacks returns UNSUPPORTED_BY_PROVIDER (exit 11). Reviews carry a scope: full (paginated) vs sample (a product-page sample — not the whole corpus).

Auth, cost & deployment

Section titled “Auth, cost & deployment”
backend keyless auth hosted/cloud-safe rough cost
serpapi ✗ API key (stdin → keyring) ✓ Free tier ~250/mo (renewing); paid metered beyond
rainforest ✗ API key (stdin → keyring) ✓ Paid, credit-based; plans from ~$25/mo
creators ✗ OAuth2 client-credentials ✓ Free API — needs an approved Associate w/ ≥10 qualifying sales / 30 days
scrape ✓ none (RIVR_SCRAPE_ENABLE=1) ✗ residential only Free; cost is your bandwidth/IP + selector upkeep
stub ✓ none ✓ Free (fixtures)

Risks

Section titled “Risks”
  • serpapi — third-party dependency; paid beyond free tier; reviews are a page sample; no browse.
  • rainforest — third-party dependency; paid per request.
  • creators — eligibility wall (ASSOCIATE_NOT_ELIGIBLE, exit 6); no review text; portal access-gated (endpoints env-overridable via RIVR_CREATORS_*).
  • scrape — Amazon ToS; bot-detection/blocking (→ BLOCKED + a persisted cooldown); fragile DOM (→ SCHEMA_DRIFT). Do not run from cloud/hosted IPs.
  • stub — not real data; offline/testing only.

Which backend should I use?

Section titled “Which backend should I use?”
  • Agent or you at home (residential): scrape (keyless, free) for modest throttled use, or serpapi (free tier) if you’d rather not scrape. This is the primary use case.
  • Hosted / cloud / datacenter: never scrape. Use creators if you’re an eligible Associate, otherwise serpapi/rainforest.
  • Need full review text: rainforest (or scrape); serpapi is a sample; creators none.
  • Need browse-node trees: creators only.

Choosing programmatically

Section titled “Choosing programmatically”
Terminal window
# keyless and usable from any host:
rivr provider list --json | jq '[.[] | select(.keyless and .hostedSafe) | .name]'
# backends that return full review text:
rivr provider list --json | jq '[.[] | select(.reviewsScope == "full") | .name]'