Pricing
Possible or not. For how much. Then you decide.
Every request gets those two answers before anything runs. Exact prices, allowances, minimum charges, and refund terms are not set yet. This page explains the structure so you can judge whether it fits before we publish numbers.
Structure
Three ways you pay.
One-time ask
A bounded collection
You approve a scope, a column set, and a row cap. The charge covers the agent’s work on that run and its result. A single table for a single question.
Unit · one approved run
Subscription
Saved collectors on a schedule
Hosted scheduling, change history, reliable exports, and a defined usage allowance or usage-based billing. Recurring runs reuse the agent’s plan, so they cost less than the first.
Unit · monthly, with a usage allowance or metered usage
Additional usage
Only with your approval
Work beyond your allowance or spending limit requires you to approve it. Nothing quietly rolls over into a larger bill.
Unit · explicitly approved overage
Later: a downloadable workflow package for running in your own environment. Whether that is a separate purchase or part of a subscription is undecided.
What moves the number
Scope, columns, volume, frequency.
Every estimate accounts for the costs below. The first run of something new costs more, because the agent has to figure the source out. Repeating it costs less. We track setup and recurring costs separately and explain any difference you see.
- Provider requests, returned records, and provider-specific billable events.
- The agent’s model usage for checking sources, planning, building the collector, and recovering from broken pages.
- Hosted compute, browser execution, and proxies where a source needs them.
- Storage, data transfer, and delivery to destinations.
- A bounded allowance for retries and failure recovery.
- Payment costs, service overhead, and a sustainable margin.
Try it
Estimate
13requests
- Rows, capped
- 300
- Listing pages
- 13
- Product-page visits
- 0
- Derived columns
- free
Illustrative units. The real estimate also covers model usage, compute, and a bounded retry allowance, and shows a price after terms are set. Nothing runs until you approve a limit.
Spending controls
A limit you set, enforced in code.
- You see an estimate or range with explicit assumptions before a paid run.
- You approve a spending limit. Changing scope or raising the limit requires new approval.
- Limits are enforced in application code at run level and account level.
- Budget is reserved per job so two jobs cannot spend the same allowance.
- Actual usage is reported after every run.
Schedules
Per run, and per month.
For a scheduled task you see an estimated cost per run and a monthly projection based on the frequency you choose. The projection states its volume assumptions. Per-run and monthly limits stop a schedule before it exceeds what you approved. How schedules work.
Not decided yet
Open commercial decisions.
- Exact prices, included usage, and minimum charges.
- Whether a sample is free or disclosed and charged.
- The policy for failed and partially completed runs.
- Refund and cancellation terms.
- Whether taxes or other charges are additional.
- Retention periods and dataset limits.
- How the code export is priced.
- Support scope and response expectations.
These decisions will be published here before any paid use. We will not present invented prices as approved terms.
Questions
Pricing questions.
Is there a sample charge?
Not decided. If a sample costs money to produce, we disclose it and ask before making any paid request.
What if a run fails or only partly completes?
The policy is not final. Whatever we decide will be written on this page before paid use. External providers may charge for partial work, and the policy has to account for that honestly.
Why does adding a column change the price?
Some columns come from the page you already read. Others need a visit to every detail page, or a second provider. We label each column as collected, derived, or extra work, and the estimate updates when you add one.
Is there an unlimited plan?
No. Every run uses provider requests, model calls, and compute that cost money. We price around a spending limit you control instead of promising unlimited use we cannot honor.
Early access
Ask for something. We will tell you if it is possible and what it would cost.
Early-access requests help us set prices against real asks, not guesses. Describe what you want, roughly how much, and how often.