Web Search API

CatchAll: recall-first web search API

Pay per result, not per token.

Abstract illustration of multiple overlapping blue outlines of envelopes fading into white background.

Web search misses over 90% of datapoints.

Most “deep research” tools only skim the surface. CatchAll flips the stack — prioritizing coverage, metadata enrichment, and precision validation before LLM reasoning.

ISSUE - 01

Limited to the first 
few pages

Ranking-first tools return 10–20 results per query. For enumeration tasks, you're left with a sample — not a dataset.

CatchALL

Goes beyond page 1. 50,000+ pages scanned per job, coverage-first. No cap on results.
Abstract 3D illustration of six progressively larger blue wireframe rectangles arranged in a staggered, overlapping pattern.
ISSUE - 02

Most data goes 
unseen

Regional press, trade publications, regulatory filings — most of the web never surfaces in standard results.

CatchALL

The sources where events actually break first — not mainstream results pages. Claude found 3 events. CatchAll found 34.
Three wireframe cube shapes in a row, each connected by blue lines, representing a digital or blockchain network concept.
ISSUE - 03

You're billed for noise

Token-based pricing punishes thorough searches. The more coverage you need, the more you pay — even for noise.

CatchALL

You only pay when an event is validated against your query. Scan 50,000+ pages, get 34 answers — pay for 34.
Illustration of blue outlined 3D cubes varying in size and grouping on a transparent background.

THE index

Not a web crawl. An event index.

Built on NewsCatcher’s proprietary web index, CatchAll turns scattered web signals into complete datasets that don’t exist anywhere else.

Built for retrieval, not browsing. Every query returns events, not pages.

>2M

Real-world events indexed daily

<hrs

To index new events

Time-series aware

The real world moves in events. Our index captures them in sequence.

how it works

From query to dataset

Step 01

Query planning

Query rewritten into multiple retrieval prompts covering different phrasings and event types.

Step 02

Retrieval

50,000+ pages scanned per job at ~10,000 pages/min.

Step 03

Clustering

Related pages grouped into single event records using the Leiden algorithm.

Step 04

Validation

Each cluster scored by LLM validators against your query. Irrelevant clusters dropped.

Step 05

Structured output

One JSON object per event, with source citations and dynamic schema.

CatchAll vs. Competitors

CatchAll finds 3x more relevant events than Exa Websets, Parallel AI FindAll, and OpenAI Deep Research.

CatchAll finds 3× more relevant events than Exa Websets, Parallel AI, and OpenAI Deep Research.

CatchAll Evaluation Charts — v3 (boxed left-rail layout)
Metric CatchAll Exa Websets Manus 1.6 Parallel AI OpenAI o3
F1 Score 0.705 0.3170.1040.1030.017
Precision 0.632 0.837 0.5940.7770.486
Recall 0.798 0.1960.0570.0550.009
Uniqueness % 85.5%18.5% 89.2% 77.6%33.7%
Query Wins 27/32 3/320/322/320/32

Best in category · Observable universe: 6,025 unique TPs across 32 queries · March 2026

CatchAll
0.705
Exa Websets
0.317
Manus 1.6
0.104
Parallel AI Core
0.103
OpenAI o3
0.017
0 0.25 0.50 0.75 1.0
F1 score — harmonic mean of precision & recall (0–1 scale)
CatchAll
79.8%4,807
Exa Websets
19.6%1,179
Manus 1.6
5.7%342
Parallel AI Core
5.5%334
OpenAI o3
0.9%53
0% 25% 50% 75% 100%
Coverage of the 6,025-event observable universe (%)

PRICING

You pay per verified result. Not per token.

Most AI search tools bill per token — compute cost that scales regardless of what you find. CatchAll charges only when an event clears validation. No match, no charge.

$0.10

per validated record

Base mode

2,000

credits free on signup

no card required

Real-time monitors

Best search is no search: enterprises don’t want a query box, they want notifications about predefined real‑world events that matter.

automation

Set it up once. Get answers continuously.

Monitor

Continuous coverage on any topic

Define a query once. CatchAll re-runs it on your schedule — hourly, daily, or weekly — and delivers only new, deduplicated events since the last run.

  • Schedule runs at any interval
  • Only new events delivered each run — no reprocessing
  • JSON output drops straight into your pipeline
Set up a monitor
Watchlist

Scope any search to companies you care about

Attach a named entity list to any job. The full web gets scanned for any mention. Results are filtered and scored by relevance to your list before delivery.

  • Built-in entity resolution across languages & jurisdictions
  • Relevance scoring per matched company (1–10)
  • One dataset, reusable across all jobs and monitors
Build your first watchlist

What this enables

Blue three-layered stacked squares icon.

Build complete datasets

Structured, verifiable records from scattered web signals.

Blue wireframe diamond-shaped geometric figure with intersecting lines inside.

Ground AI agents in complete data

Agents on incomplete search miss 80–90% of relevant events.

Blue circular icon with two curved lines dividing the circle into three parts, resembling a stylized globe or sphere.

Eliminate compliance blind spots

79.8% observable recall across regulatory, risk, and supply chain events.

Abstract blue symmetrical shape with three curved arches stacked vertically.

Structured output

Events extracted, deduplicated, and cited. Structured JSON, ready to use.

FAQ

Is CatchAll the same as the News API?

No. They are separate products with separate API keys and endpoints. CatchAll is a web search API focused on real-world events. News API is a general news search API for enterprise customers only.

Is CatchAll free to use?

Yes, there's a free tier. You get 2,000 credits upon sign up + 10 searches per month. Beyond that, you need to upgrade to a paid plan.

How does billing work?

You're billed per valid record returned, not per API call. Zero results means zero charge. Lite mode is billed per job at a fixed 100 credits, returning up to 100 results.

How long does a search take?

Lite mode returns results in seconds. Base mode typically takes around 15 minutes. Jobs show as "pending" while processing.

How do I write a good query?

Use the formula: event type + time window + optional scope. Example: "Series A funding announcements in Europe over the past 30 days." More specific queries return better results.

Does CatchAll support non-English sources?

Yes. CatchAll indexes international sources and may return results in multiple languages. Add language filters in the context field for English-only results.

Monitors is a feature of CatchAll — it reruns your search automatically on a schedule.

How is this different from running a cron job myself?

A Monitor handles the date windowing, deduplication, and delivery. You get a clean feed of new records — no infrastructure to maintain.

What's the minimum schedule interval?

Hourly.

Can I attach a Watchlist to a Monitor?

Yes. Attach a dataset with your entity list to scope results to specific companies. Watchlist scoring applies to every run.

Can I pause a monitor without deleting it?

Yes. Disable pauses execution without removing configuration or history. Re-enable to resume on the same schedule.

Watchlist is a feature of CatchAll — it scopes your search to a list of companies you choose.

Is this a news alert tool?

No. Alerts send links. Watchlist returns deduplicated event records, filtered to your companies and scored for relevance.

What counts as a "mention"?

Any page where a tracked entity is confirmed relevant by the validator. Entity resolution, not keyword matching.

How is relevance scored?

LLM validator scores each result 1–10 against your entity list. 7+ is typically high relevance.

Can I use the same Watchlist across multiple searches?

Yes. One list across all jobs and monitors. Update once.

Does it support non-English companies?

Yes. Name variants, aliases, and non-English spellings handled by default.

See what you've been missing.

You can clean data. You can't materialize missing data.

Start free
2,000 credits included
Book a Demo
|
GDPR compliant
|