Pricing

• A call without providers in academic costs one flat price, whichever engine answers. Every other call pays the answering engine's list price × 1.5, at least $0.0001 a call.
• Pin an engine with providers and you pay that engine's rate.
• Failed, timed-out and empty calls are not charged.
• A repeat of the same request shortly after is answered from the stored result at the same price, with no wait; send fresh: true to search again.
• Before a call we hold the most it can cost, then refund the difference the moment it ends. Without providers, that is the largest max among the engines the router may try, or in academic its flat price.
• When one call asks several engines, you pay for each engine that returned results.
• Auto routing and relevance reranking add no charge. Auto never picks deep. It picks social (from $0.00282 a call) only on a platform cue: "tweet", "Twitter", "on X", an x.com link or an @handle for X; "on tiktok" or "tiktok videos"; "instagram reels" or "on instagram"; "youtube videos", "youtube comments" or "on youtube". A named platform is searched alone, one call. Code such as @types/node, @media or Twitter cards, and questions about the platforms themselves (the YouTube API, TikTok's stock), do not count.
• Credits never expire while your account is open.
• Top up any whole amount from $5 to $1,000. No subscription.

What things cost

One price per mode

A flat price is the same whichever engine answers; pin an engine with providers and you pay that engine's rate. Failed calls are free.

Price per engine

What a call pinned to one engine with providers pays, and what a call without providers pays where its mode has no flat price. Typical is a call that returns 10 results. Max is the most a call pinned to this engine can cost. Without providers we hold the largest max among the engines the router may try, or in academic its flat price, and refund the rest. The hold under each table is for the default engine order; when an engine is down the router skips it and the hold follows the engines it will try. Prices are in US dollars and include our margin.

general · news · academic · people · deep · extract · similar · scholar · patents · images · places · social · community

Questions

Why does one mode have several prices?

Each engine has its own list price. Name the engines you want in providers and you pay their rates. Leave it out and the router tries engines in order until one returns results: in a flat-priced mode you pay the flat price whichever answered, otherwise you pay for the engine that answered.

What is the maximum per call?

Some engines bill by what a call used: results beyond the first 10, PDF pages, tokens or X posts read. Before the call we hold the maximum from your balance; afterwards you pay what the engine used and the rest comes back. A call never costs more than its maximum. Max is the most a call pinned to this engine can cost. Without providers we hold the largest max among the engines the router may try, or in academic its flat price, and refund the rest. Each mode lists that hold under its table. For example, an extract without providers holds $0.1575 and usually costs $0.00075, so a balance below $0.1575 gets 402 insufficient_credits.

What happens when a call fails?

Nothing is charged. Failed, timed-out, rejected and empty attempts are free, and the hold is returned at once. The response lists every attempt and what each cost.

What does unconfirmed mean?

The vendor does not publish that figure plainly, so we could not check it against their own page. We price it at the top of our reading and lower it once it is confirmed.

Can I use my own provider keys instead?

Yes. search-cli is open source and calls the providers with your own keys at their prices, with no account here. Log in when you want the engines you have no key for.

Do credits expire?

No, not while your account is open. Every top-up gets an invoice you can download as a PDF.

Compared with going direct

• metasearch.sh vs Tavily
• metasearch.sh vs Exa
• metasearch.sh vs SerpApi