SscoutStart building

// web intelligence api

The web, wired for your agents.

Scout is a web-intelligence API for builders. One set of endpoints to search, extract, research, enumerate, and monitor the live web - returning structured data with real citations, not a wall of links. No SDK and no setup: send an HTTP request and your agent is online.

curl https://forager-hirargb.fly.dev/v1/search \
  -H "content-type: application/json" \
  -d '{"queries": ["best vector databases"], "limit": 5}'

// the toolkit

Six endpoints. One base URL.

Most teams assemble web access from half a dozen vendors and a pile of glue code. Scout is six endpoints that share one key and one balance - use one, use all six, compose them freely.

01/v1/search

Search

Rank the live web for any query. Several queries at once, domain filters, freshness windows, and clean excerpts your model can quote.

02/v1/extract

Extract

Turn any URL into clean markdown - or, with an output schema, into the exact JSON shape you defined. No parsing, no cleanup.

03/v1/task

Task

Hand off a research objective in plain language. Get a structured answer with a per-field citation basis: sources, excerpts, confidence.

04/v1/findall

FindAll

Enumerate every entity matching a description as typed rows. Extend a finished run to find more, enrich it to add columns.

05/v1/monitors

Monitors

Watch a query on a schedule. When the web changes, Scout delivers a webhook with the new results - hourly, daily, or weekly.

06/v1/chat/completions

Chat

An OpenAI-compatible chat endpoint grounded in live web results. Point an existing client straight at it.

// search

Search the live web.

One request, several queries, merged and de-duplicated. Results come back as clean URLs, titles, and excerpts - ready to cite.

scout ~ search...
$ POST /v1/search
{ "queries": ["open-source vector databases"], "limit": 3 }
# 200 ok . 1 credit
{ "results": [
{ "url": "qdrant.tech", "title": "Qdrant" },
{ "url": "weaviate.io", "title": "Weaviate" },
{ "url": "milvus.io", "title": "Milvus" }
] }
_

// extract

Extract into your own schema.

Hand Scout a JSON schema and it coerces the page into exactly that shape before it reaches your agent. The web, already typed.

scout ~ extract...
$ POST /v1/extract
{ "urls": ["anthropic.com/company"],
"output_schema": { "ceo": "string", "founded_year": "integer" } }
# 200 ok . structured . 4 credits
{ "data": { "ceo": "Dario Amodei", "founded_year": 2021 } }
_

// task

Hand off the research.

Describe an objective in plain language. Scout searches, reads, and reasons across sources, then returns an answer with its citation basis.

scout ~ task...
$ POST /v1/task
{ "task": "Who are the main competitors to Anthropic?" }
# run-9c4f1a . completed . citation basis attached
{ "answer": "OpenAI, Google DeepMind, Mistral, xAI",
"confidence": "high" }
_

// why scout

Most web APIs hand your agent a firehose. Scout hands it answers.

A list of links is where the work begins, not where it ends. Scout searches, reads, reasons, and structures - so you ship an agent instead of a scraping pipeline. Three decisions shaped it.

01

One API, not five

Search, extract, research, enumerate, monitor, and chat behind one base URL, one key, one balance. No pipeline of a scraper, a search vendor, and a parser to keep alive.

02

Structured by default

Pass a JSON schema, get exactly that shape back. Every Task answer ships with a citation basis - the sources, the excerpts, the confidence - so your agent can show its work.

03

Built to ship

No SDK to install, no key to wait for - plain HTTP and JSON. Long jobs run async: poll them, stream them over SSE, or take a webhook when they finish.

// pricing

Priced in credits. No token math.

Every call spends a fixed, predictable number of credits - a search is one, a deep research task a handful. Start with 1,000 free credits a month, no card. The pricing page has an estimator that turns your expected volume into a real monthly number.

// get started

Online in three steps.

01

Read the docs

Endpoint reference, quickstart, and copy-paste examples for every call.

Open docs ->
02

Send a request

Point a curl at the base URL. No key required to start - the free tier is open.

Quickstart ->
03

Pick a plan

Estimate your monthly cost in credits and move to a plan when you outgrow free.

See pricing ->

// ready

Give your agents the whole web.

Free to start, no card, online in a single request. Point Scout at the live web and build the agent you actually wanted.