# Spend less on FetchAPI: a cost guide

Every answer from `https://fetchapi.co/api/gw/<provider>/<path>` carries the upstream cost of that call in
`meta.cost_usd`. Your spend by endpoint, repeat requests and cheaper same-job routes are on
https://fetchapi.co/account under "Where your money goes". An agent can read the same through the MCP tool `get_usage`.

Across all keys in the 30 days to 2026-10-08, 0.4% of calls made 28% of the bill, and repeat requests made 18%.
The long version, with the numbers and real calls: https://fetchapi.co/blog/spend-less-on-data-api-calls

## The rules

1. **Ask for what you will read.** Per-row prices are the expensive ones. DataForSEO LLM Mentions search is $0.10 a
   call + $0.001 a row: `limit: 1000` costs up to $1.10, `limit: 10` costs $0.11 and still returns `total_count`.
   Ranked Keywords is $0.012 + $0.00012 a keyword. Set `limit` / `depth` to the rows you need.
2. **Use the summary endpoint for counts.** `ai_optimization/llm_mentions/aggregated_metrics` answers "how many
   mentions" for about $0.10 with no row charge. Pull rows only for targets that pass your threshold.
3. **Cache.** Store each answer with its fetch time; reuse it until the data can have changed (a day for profiles,
   a month for monthly metrics). A repeated identical request costs the same as the first one.
4. **Queue what can wait.** DataForSEO Google results: $0.0006 a page in the standard queue (`serp/google/organic/task_post`,
   then `task_get`), $0.002 live. Same page, 3.3x cheaper.
5. **Pick the cheapest route for the job.** The same job often has several endpoints. Find them with
   `search_endpoints("<job>")` over MCP or https://fetchapi.co/catalog and compare `usd_per_call`. Listing a TikTok
   user's posts is about $0.00002 a call on one provider and $0.002 on another.
6. **Watch variable prices.** LLM Responses (ChatGPT, Claude, Gemini) add the model's token charge: shorter prompts and
   capped answers cost less.
7. **Retry only what can succeed.** Failed calls are free (https://fetchapi.co/terms), but a bad input fails again. Retry
   timeouts and 5xx with back-off; fix 4xx inputs.

## Before a big job

1. Run 10 inputs and read `meta.cost_usd` for each.
2. Multiply by your volume. The gateway's daily and monthly caps protect you; your key's cap is on /account.
3. Check /account after the first day for the top endpoints and any cheaper same-job route.

## Reference

- Price per call for every endpoint: https://fetchapi.co/gateway-catalog.json (`usd_per_call`, `price_source`)
- Per-provider docs: https://fetchapi.co/docs/<provider>.md
- Charges, caps and what is free: https://fetchapi.co/terms