NewYour Exa or Tako SDK, up to 7× cheaper.

Custom index

Search the pages you choose.

Add a sitemap, a feed or a list of URLs. We read them, re-read them every day, and your agent searches them with the same API: every result with the probability that it’s the right one, plus the answers you ask for, like price or stock.

POST /v1/indexes$2.00 / 1k URLs a month

idx_8f3k2m9qtrail running shoes under $120

Catalog

active

  • 1,284ready
  • 12pending
  • 3blocked
  • sitemap.xml · checked 14 min ago
  • Every URL re-read daily
  • Ridgeline Trail 4 · trail running shoe

    shop.example/p/ridgeline-trail-4

    $109InStockin_stock 0.98
    0.96
  • Canyon Grip GTX · waterproof trail shoe

    shop.example/p/canyon-grip-gtx

    $119InStockin_stock 0.95
    0.92
  • Summit Racer 2 · lightweight trail racer

    shop.example/p/summit-racer-2

    $124OutOfStockin_stock 0.03
    0.41
Sample output

Up and running in minutes

  1. 1

    Add your pages

    Send a list of URLs, a sitemap or an RSS feed, up to 1,000 URLs per request. We read every public page right away.

    POST /v1/indexes/{id}/urls

  2. 2

    It stays fresh on its own

    We re-read every URL daily and check sitemaps and feeds every hour, so new pages show up by themselves.

    daily · hourly

  3. 3

    Your agent searches

    Add index to your search, from the API or the MCP server. Same modes, same prices.

    index: "idx_…"

You pick the pages. We do the rest.

  • 01

    Nothing to build

    No crawlers, no embeddings, no cron jobs. We read your pages, re-read them daily and pull out price, stock, brand and SKU when the page declares them.

  • 02

    Relevance you can trust

    Every result carries the probability that it’s the right one. 0.9 means yes, 0.5 means unsure, and we say so. Set a threshold and move on.

  • 03

    Answers, not just links

    Ask what you need to know, like whether size M is in stock or the price changed, and get the answer on every result, in the same call.

One index, many uses

Pick your sector to see what you add and what your agent gets back.

Answers straight from your catalog

A shopper asks “do you have the Alpine rain jacket in size M?”. Your agent searches your product pages and replies with the link, the price and whether it’s in stock.

What you add

  • Your product sitemap
  • Competitor product pages

What your agent gets

  • The right product, with its link
  • Whether size M is in stock
  • Price and availability as data
Sample output
Alpine rain jacket in size M
  • Alpine Rain Jacket · men’s

    shop.example/p/alpine-rain-jacket

    $189InStocksize_m_in_stock 0.94
    0.97
  • Alpine Rain Jacket · women’s

    shop.example/p/alpine-rain-jacket-w

    $189InStocksize_m_in_stock 0.08
    0.84
  • Storm Shell 2 · rain jacket

    rival.example/storm-shell-2

    $159OutOfStocksize_m_in_stock 0.03
    0.46
See the request · POST /v1/search
{
"query": "Alpine rain jacket in size M",
"index": "idx_8f3k2m9q",
"mode": "fast",
"questions": {
"size_m_in_stock": {
"type": "boolean",
"instructions": "Is size M in stock?"
}
}
}

Why not a vector database or your site search?

Both are great at what they were made for. A custom index is made for agents that have to act on the answer.

Why not a vector database or your site search?Vector databaseSite searchCustom index
What you get backHow similar the text is, not whether it answersKeyword matches and your boostsThe probability that the page answers the query
Who reads the pagesYou: fetch, chunk and embedYou: push records or set up a crawlerWe do, from your URLs, sitemaps and feeds
Staying freshYou re-embedYou reindexOn its own: every day, sitemaps every hour
Price, stock and other answersAnother step you buildFilters you define up frontOn every result, in the same call
Made forRetrieval inside your appPeople typing in a search boxAgents, over the API and MCP

Need millisecond retrieval over millions of records, or a search box for your shoppers? Keep those. A custom index is for when your agent needs a reliable answer from pages you choose.

API

One more parameter on your search

If you already search with typesearch, it’s one more parameter: POST /v1/search with index, in plain JSON. The MCP server’s search_news takes it too, so your agent searches your pages and the news with the same tool.

# 1. Create an index
curl https://api.typesearch.ai/v1/indexes \
-H "Authorization: Bearer $TYPESEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "name": "Catalog" }'
# 201 Created · { "id": "idx_8f3k2m9q", "object": "index", … }
 
# 2. Add a sitemap (or up to 1,000 URLs, or a feed)
curl https://api.typesearch.ai/v1/indexes/idx_8f3k2m9q/urls \
-H "Authorization: Bearer $TYPESEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "sitemap": "https://shop.example/sitemap.xml" }'
# 202 Accepted · read right away, then every day
 
# 3. Search it
curl https://api.typesearch.ai/v1/search \
-H "Authorization: Bearer $TYPESEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "query": "trail running shoes under $120", "index": "idx_8f3k2m9q", "mode": "fast" }'

Your pages, in good hands

  • Only public pages

    We read what anyone can open: nothing behind a login, cookies or a paywall. Private addresses and internal networks are rejected.

  • We respect every site

    We follow robots.txt and the same opt-out rules as the news index. If a page doesn’t want to be read, it comes back as blocked.

  • Yours alone

    Your index belongs to your organization. No other customer can search it.

  • Never used to train models

    Not your URLs, not the pages we read for them, not your queries.

Pricing

Pay only for the URLs you keep

$2.00/ 1k URLs a month

An index with 100 URLs costs $0.20 a month; one with 10,000, $20. Billed daily from the same prepaid credit, with re-reading included. Searching it costs the same as any search: $1.00 per 1,000 in ultra, $1.40 in fast, $2.20 in normal and $5.60 in deep.

URLs per plan

  • Pay as you goUp to 100 URLs · 3 indexes
  • Monthly credit planUp to 10,000 URLs · 50 indexes
  • EnterpriseUnlimited
More than 100 URLs? They come with the monthly credit plan, from $50 a month.

Searches over your index

  • ultra$1.00
  • fast$1.40
  • normal$2.20
  • deep$5.60

/ 1k searches

Custom index questions

What can I put in an index?

Any public page: products, investor relations, press releases, official gazettes, reviews, documentation. Add URLs directly (up to 1,000 per request), or give us a sitemap or an RSS or Atom feed and we take the URLs from it.

How fresh is it?

Every URL is re-read once a day. Sitemaps and feeds are checked every hour, and new or updated URLs are read right away. Each URL shows when it was read, when it last changed and when it will be read next.

How much does it cost?

$2.00 per 1,000 URLs a month, billed daily from your prepaid credit: 100 URLs cost $0.20 a month, and 10,000 URLs on the monthly plan cost $20. Re-reading is included, and searching your index costs the same as any search in that mode.

How many URLs can I add?

Up to 100 URLs in 3 indexes pay as you go, and up to 10,000 in 50 indexes with the monthly credit plan. Enterprise has no limit. URLs past your limit come back as over_limit and are not added.

What happens if my credit runs out?

The index is paused: your URLs stay, but they aren’t re-read until you top up. Then it picks up where it left off.

Can my agent use it from the MCP server?

Yes. search_news takes an optional index parameter: pass your idx_… and it searches your index instead of the news, with the same key and prices.

Do you read pages that block crawlers?

No. We honor robots.txt and the same opt-out rules as the news index. A URL we can’t read comes back as blocked, and we try it again a week later.

Can I search my index and the news together?

Each request searches one or the other, and a request with index takes one query. Send the same query twice, with your index and without it, to see your pages next to the press coverage.

Point your agent at your own pages.

Create an index, add a sitemap and run your first search in minutes. Up to 100 URLs pay as you go, 10,000 on the monthly plan.