Scry, programmable internet search w/ congestion pricing

live
CategorySEO and detection
Rate cardpricing page
Last check24 Sept 2026

Rate card history

Pricing lines captured on each sweep. Added and removed lines appear in the changelog the day they change.

Observed 24 Sept 2026
  • Every account starts with $5 of credit and no card.
  • Slack While nobody is waiting for a slot, a query costs $0. The response still reports burden_nanodollars , what the run metered, beside spend_nanodollars , what you paid.
  • Meter The meter is the larger of slot-seconds held and resources used (CPU-seconds, reads at $0.000625 per logical GB and about $0.0014 per physical GB, memory byte-seconds), plus a $0.00001 floor. Rent is c times the la
  • Ceiling max_multiplier is the highest c you accept, 5 when omitted. It is the most you pay and your place in the line: you pay the posted price, never your ceiling. Above it the call is refused as priced_out with the pri
  • Line The next slot goes to the highest max_multiplier in the line, then to whoever holds the fewest, weighted by class: at the default ceiling, hold nothing and you are next. patient waits until the price fits. priority
  • Limits max_seconds , budget_nanodollars and max_exposure_nanodollars are ceilings you set, never fees, and an unspent budget costs nothing. In the week ending 2026-09-10 the median statement metered $0.010 and the 99th p
  • Also metered Response bytes past a free daily band, in every regime. Agents without an account prepay over x402, $0.05 per second of the time box they declare. zero_retention pays 10 times the metered component, $0.25 mi
Observed 23 Sept 2026
  • Every account starts with $5 of credit and no card.
  • Slack While nobody is waiting for a slot, a query costs $0. The response still reports burden_nanodollars , what the run metered, beside spend_nanodollars , what you paid.
  • Meter The meter is the larger of slot-seconds held and resources used (CPU-seconds, reads at $0.000625 per logical GB and about $0.0014 per physical GB, memory byte-seconds), plus a $0.00001 floor. Rent is c times the la
  • Ceiling max_multiplier is the highest c you accept, 5 when omitted. It is the most you pay and your place in the line: you pay the posted price, never your ceiling. Above it the call is refused as priced_out with the pri
  • Line The next slot goes to the highest max_multiplier in the line, then to whoever holds the fewest, weighted by class: at the default ceiling, hold nothing and you are next. patient waits until the price fits. priority
  • Limits max_seconds , budget_nanodollars and max_exposure_nanodollars are ceilings you set, never fees, and an unspent budget costs nothing. In the week ending 2026-09-10 the median statement metered $0.010 and the 99th p
  • Also metered Response bytes past a free daily band, in every regime. Agents without an account prepay over x402, $0.05 per second of the time box they declare. zero_retention pays 10 times the metered component, $0.25 mi
Observed 22 Sept 2026
  • Every account starts with $5 of credit and no card.
  • Slack While nobody is waiting for a slot, a query costs $0. The response still reports burden_nanodollars , what the run metered, beside spend_nanodollars , what you paid.
  • Meter The meter is the larger of slot-seconds held and resources used (CPU-seconds, reads at $0.000625 per logical GB and about $0.0014 per physical GB, memory byte-seconds), plus a $0.00001 floor. Rent is c times the la
  • Ceiling max_multiplier is the highest c you accept, 5 when omitted. It is the most you pay and your place in the line: you pay the posted price, never your ceiling. Above it the call is refused as priced_out with the pri
  • Line The next slot goes to the highest max_multiplier in the line, then to whoever holds the fewest, weighted by class: at the default ceiling, hold nothing and you are next. patient waits until the price fits. priority
  • Limits max_seconds , budget_nanodollars and max_exposure_nanodollars are ceilings you set, never fees, and an unspent budget costs nothing. In the week ending 2026-09-10 the median statement metered $0.010 and the 99th p
  • Also metered Response bytes past a free daily band, in every regime. Agents without an account prepay over x402, $0.05 per second of the time box they declare. zero_retention pays 10 times the metered component, $0.25 mi
Observed 21 Sept 2026
  • Every account starts with $5 of credit and no card.
  • Slack While the machines are at most half occupied, a query costs $0. The response still reports burden_nanodollars , what the run meters at full engagement, beside spend_nanodollars , what you paid.
  • Meter Past half occupancy ρ the meter engages by e = ((ρ − ½) / ½)² : 0 at half, a tenth at two thirds, 1 at a full box. Each second of a query pays that second's e. The meter is the larger of slot-seconds held and resou
  • Price A controller reads occupancy every 150 ms and posts a congestion multiplier c, in the shape of Ethereum's base fee (EIP-1559). Its target is one free slot. Above target c rises at most 0.35% per epoch, doubling in
  • Ceiling max_multiplier is the highest c you accept, 5 when omitted. It is consent, never a bid: you pay the posted price, not your ceiling. Above it the call is refused as priced_out with a forecast of when the price wil
  • Line The next slot goes to whoever holds the fewest, weighted by class: hold nothing and you are next. patient waits until the price fits. priority doubles your weight at double rent, at least $0.001. Every account runs
  • Limits max_seconds , budget_nanodollars and max_exposure_nanodollars are ceilings you set, never fees, and an unspent budget costs nothing. In the week ending 2026-09-10 the median statement metered $0.010 at full engage
  • Also metered Response bytes past a free daily band, in every regime. Agents without an account prepay over x402, $0.05 per second of the time box they declare. zero_retention pays 10 times the metered component, $0.25 mi

Signal history

24 Sept 2026
rate cardScry, programmable internet search w/ congestion pricingupdated its pricing page
  • Page content changed but no pricing lines were added or removed. Possibly terms or plan descriptions.
23 Sept 2026
rate cardScry, programmable internet search w/ congestion pricingupdated its pricing page
  • Page content changed but no pricing lines were added or removed. Possibly terms or plan descriptions.
22 Sept 2026
rate cardScry, programmable internet search w/ congestion pricingchanged its pricing page
  • Removed: Slack While the machines are at most half occupied, a query costs $0. The response still reports burden_nanodollars , what the run meters at full engagement, beside spend_nanodollars , what you paid.
  • Removed: Meter Past half occupancy ρ the meter engages by e = ((ρ − ½) / ½)² : 0 at half, a tenth at two thirds, 1 at a full box. Each second of a query pays that second's e. The meter is the larger of slot-seconds held and resou
  • Removed: Price A controller reads occupancy every 150 ms and posts a congestion multiplier c, in the shape of Ethereum's base fee (EIP-1559). Its target is one free slot. Above target c rises at most 0.35% per epoch, doubling in
  • Removed: Ceiling max_multiplier is the highest c you accept, 5 when omitted. It is consent, never a bid: you pay the posted price, not your ceiling. Above it the call is refused as priced_out with a forecast of when the price wil
  • Removed: Line The next slot goes to whoever holds the fewest, weighted by class: hold nothing and you are next. patient waits until the price fits. priority doubles your weight at double rent, at least $0.001. Every account runs
  • Removed: Limits max_seconds , budget_nanodollars and max_exposure_nanodollars are ceilings you set, never fees, and an unspent budget costs nothing. In the week ending 2026-09-10 the median statement metered $0.010 at full engage
  • Added: Slack While nobody is waiting for a slot, a query costs $0. The response still reports burden_nanodollars , what the run metered, beside spend_nanodollars , what you paid.
  • Added: Meter The meter is the larger of slot-seconds held and resources used (CPU-seconds, reads at $0.000625 per logical GB and about $0.0014 per physical GB, memory byte-seconds), plus a $0.00001 floor. Rent is c times the la
  • Added: Ceiling max_multiplier is the highest c you accept, 5 when omitted. It is the most you pay and your place in the line: you pay the posted price, never your ceiling. Above it the call is refused as priced_out with the pri
  • Added: Line The next slot goes to the highest max_multiplier in the line, then to whoever holds the fewest, weighted by class: at the default ceiling, hold nothing and you are next. patient waits until the price fits. priority
  • Added: Limits max_seconds , budget_nanodollars and max_exposure_nanodollars are ceilings you set, never fees, and an unspent budget costs nothing. In the week ending 2026-09-10 the median statement metered $0.010 and the 99th p

Back to all tools