Forecasting news infrastructure

Backtest on the news as it existed then.

Reconstruct exactly what was knowable at any point in time, validate your model against it, then deploy on the same licensed source in real time. No future-data leakage.

Explore forecasting data Read the backtesting guide

cli.sh

$ asknews research "Determine what was knowable about Nvidia earnings on 2024-02-21"

Markets42 Sources

Traders price in a longer hold after inflation surprise

3 min agoFOMC CPI Treasuries

Finance61 Sources

Rate-path expectations diverge across the curve

2 min agoECB BoJ Gilts

Elections88 Sources

Polling averages shift after late-cycle debate

22 min agoPennsylvania Michigan Arizona

Commodities190 Sources

Supply forecasts revised as export quotas tighten

1 day agoBrent OPEC Copper

Geopolitics430 Sources

Sanctions package reshapes trade-flow expectations

6 days agoEU Sanctions Tariffs

Markets42 Sources

Traders price in a longer hold after inflation surprise

3 min agoFOMC CPI Treasuries

Finance61 Sources

Rate-path expectations diverge across the curve

2 min agoECB BoJ Gilts

Elections88 Sources

Polling averages shift after late-cycle debate

22 min agoPennsylvania Michigan Arizona

Commodities190 Sources

Supply forecasts revised as export quotas tighten

1 day agoBrent OPEC Copper

Geopolitics430 Sources

Sanctions package reshapes trade-flow expectations

6 days agoEU Sanctions Tariffs

Used in the field

From the UK Cabinet Office and Frontier AI Labs to organizations across insurance, intelligence, private equity, and hedge funds.

Built on trusted, licensed sources

The problem

Your backtest is only as honest as its archive.

Scraped archives get re-edited, re-dated, and back-filled. The moment tomorrow’s information reaches yesterday’s test, the result stops meaning anything.

Future-data leakage

Re-dated and updated articles quietly contaminate the evaluation.

Thin scraped archives

Paywalled reporting is missing exactly where the signal was.

Two different sources

Testing on one corpus and deploying on another invalidates the result.

The fix is an archive that preserves time: AskNews.

Why forecasters switch

A backtest you can actually defend.

Temporal integrity, licensed depth, and one consistent corpus from evaluation through live deployment.

0: future-data leakage in bounded queries
200M+: historical articles to test against
1: source for backtest and production
500K+: new articles indexed every day

See how time-bounding works

Reconstruct who and what was connected to an event — as of any date.

Rate Decision

Christine Lagarde

European Central Bank

Frankfurt

Euro Area

Bond Markets

Deutsche Bank

IMF

Kristalina Georgieva

PeoplePlacesOrganizations

Mentions over timeVolume · Sentiment

Forecasting-grade infrastructure

One source for the backtest and the deployment.

Validate against the historical record, then run live on the same licensed feed — so nothing about your data layer changes when the model goes to production.

Time-bounded retrieval

Query the corpus exactly as it stood at any timestamp.

Historical News

Years of licensed reporting, consistently dated.

News API

The same endpoint in backtest and in production.

High-volume access

Sweep thousands of resolution windows without rate-limit gymnastics.

Knowledge Graph

Entity-level features that stay stable across the archive.

MCP/CLI for Agents

Wire the archive straight into agentic research loops.

See all capabilities

Enrichments

400% more enrichments/extractions than competitors.

Every AskNews datapoint arrives with the context models need—not a raw article your team still has to process. Real-time news access responds in under 100 ms.

Under 100 ms real-time response

Technology42 sources · 18 sec ago

Chipmakers expand US capacity as AI infrastructure demand accelerates

AskNews resolves the article into structured, model-ready context as it arrives.

Wikidata entitiesNVIDIA · Q76

Geo-coordinates37.7749° N, 122.4194° W

BiasCenter · 0.08

SentimentPositive · 0.74

Reporting VoiceAnalytical

Graph RelationshipsNVIDIA → TSMC → Arizona

Key PointsToken optimized · 3 points

One toolbox

Every source your research needs. One API.

Reach paywalled news, the open web, X, Wikipedia, Google, Reddit, and Podcasts through one consistent AskNews toolbox—without stitching together seven separate integrations.

Bring licensed reporting and the broader web into the same research workflow, then let AskNews retrieve, connect, and structure the evidence.

Paywalled news

Open web

Wikipedia

Google

Podcasts

Research traversalself-learning

Competitors

QDepth 1

One pass. One result set. Research stops here.

DeepNews

Depth 1

NewsPaywalled news

Depth 2

NewsPaywalled newsWebOpen web

Depth 3

NewsPaywalled newsWebOpen webXX

Depth 4

NewsPaywalled newsWebOpen webXXWikiWikipedia

Iterates, learns, branches, and verifies

Depth

Research that keeps going when others stop.

DeepNews performs self-learning iterative research across paywalled news, open web content, X, Wikipedia, and much more, to reach never-before-seen research depth and breadth.

Built for research engineers

Reproducible from the first sweep.

CLI
REST API
Python & TypeScript SDKs
MCP for agents
Comprehensive docs

View documentation

Getting started

From query to production in just three steps.

01
Set your as-of date
Bound every query to what was knowable then.
02
Sweep the archive
Run resolution windows across years of coverage.
03
Deploy live
Point production at the same licensed source.

How do we compare?

Most archives cannot tell you what was knowable when.

For forecasting, the question is not how much history you can reach. It is whether that history is honest about time.

Approach	The problem	AskNews
Scraped archives	Re-dated and back-filled after the fact	Timestamps preserved as published
Generic AI search	Returns today’s view of a past event	Bound every query to an as-of date
Traditional news APIs	Shallow history, no paywalled depth	Licensed archive spanning years
Separate test and live feeds	Backtest never matches production	One source across both

Customer proof

Built for systems that have to be right about time.

Used by forecasting and investment teams to test against the historical record and monitor live events on the same news layer.

Nathan AckermanPresident & CEO at Riley Risk Inc.

"Our analysts are saving valuable time doing geopolitical and security risk analysis for our clients with AskNews' impressive knowledge graphs and forecast tools."

Claudia TietzeCEO & Creative Technologist at Farallon, LLC

"You solve a lot of problems that plague other news-based intelligence product. I put the same queries into Newsplunker as my other OSINT tools. Newsplunker does real research, the other tools struggle."

Tom LiptayDirector of Forecasting at Metaculus

"The majority of winners of the Q3 forecasting tournament used AskNews. That's why our staff converted all of our in-house bots from Perplexity to AskNews for the Q4 tournament."

David SchmeltzlePresident at BizBudding

"When we needed a news API for our sports analytics platform ProMatchups, AskNews was the clear choice. The quality of the data and the ease of use is unmatched."

Marc GraczykCEO at Infinite Games

"Our use-case demands the highest quality data without any sacrifices. AskNews takes this 10 steps further."

Alessandro AmaroCo-founder & CPO at Desia

"We're going into production with AskNews. Your data has much better results quality and latency than the competition."

Arash AminiPostdoctoral Fellow, UAutonomy Lab, University of Texas at Austin

"Our research into misinformation detection with CrediRAG required the highest quality data source. We landed on AskNews and it far exceeded our expectations."

Carolyn MeinelForecasting Expert & President at BestWorld

"After weeding out the most important components for forecasting accuracy, we've identified AskNews as a vital asset and are now quickly climbing the leaderboard in the Metaculus Q4 tournament."

Matthew GranadeLLM developer

"AskNews is a core component of my application, providing high-quality news context when it matters most."

Phil GodzinStaff Software Engineer at JW Player

"In order to win the Q4 Metaculus forecasting tournament, I used great context engineering (AskNews) to make sure the AI used the most up-to-date information."

Joe SegalRAG Specialist

"I would like to thank AskNews for helping my application achieve incredible business insights."

Krasimir AtanasovValue Hunter

"I really liked the AskNews API and the search capabilities."

BohdanNewsplunker enthusiast

"I really love your product ."

Janna LuEvaluating LLMs on Real-World Forecasting Against Expert Forecasters

"An LLM does better when fed news articles from AskNews over Perplexity, mostly because the news articles are more up-to-date. Perplexity sometimes would return news articles that are a few years old, which could be helpful but often confuses the model."

PascaleLeidos

"This tool is saving me countless hours for research. I wish I could give it a second endorsement!"

99.9%: Service SLA
200,000+: Global sources
500K+: New articles per day
200M+: Historical articles

Pricing

Find the right plan for you

Access trusted AI-powered news intelligence in the way that suits your needs.

No subscription

Pay as you go

Start free with included API usage — add a card, no initial top-up required.

$0/month

25 credits included
150,000 Lite tokens included
150,000 Standard tokens included

Personal API

Pro

A low-cost plan for personal API access and deeper research.

$7.99/month

500 credits included per month
2M bonus tokens included per month
100 bonus credits per month
20% lower credit recharge rates

Deep research

Spelunker

For teams and power users who need more depth and volume.

$250/month

20,000 credits included per month
10M bonus tokens included per month
3,333 bonus credits per month
40% lower credit recharge rates
Double rate limit
Exclusive Tools (X, Charts, Websearch)

Professional

Analyst

Professional API capacity for production research workflows.

$1,000/month

110,000 credits included
45M bonus tokens included
10,000 bonus credits
60% lower credit recharge rates
4x rate limit
Exclusive Tools (Reddit)
Premium data

Commercial scale

Enterprise

Commercial licensing, custom usage, support, and deployment terms.

As low as $0.0015/credit
Rate-limits as high as 5000 RPS
Enterprise support and SLAs

Compare all plans and features

Forecast with confidence

Validate on history you can actually trust.

Join the forecasting labs and event-risk teams running backtests and live systems on the same licensed news layer.

Explore forecasting data

cli.sh

$ asknews research "Determine what was knowable about Nvidia earnings on 2024-02-21"

Markets42 Sources

Traders price in a longer hold after inflation surprise

3 min agoFOMC CPI Treasuries

Finance61 Sources

Rate-path expectations diverge across the curve

2 min agoECB BoJ Gilts

Elections88 Sources

Polling averages shift after late-cycle debate

22 min agoPennsylvania Michigan Arizona

Commodities190 Sources

Supply forecasts revised as export quotas tighten

1 day agoBrent OPEC Copper

Geopolitics430 Sources

Sanctions package reshapes trade-flow expectations

6 days agoEU Sanctions Tariffs

Markets42 Sources

Traders price in a longer hold after inflation surprise

3 min agoFOMC CPI Treasuries

Finance61 Sources

Rate-path expectations diverge across the curve

2 min agoECB BoJ Gilts

Elections88 Sources

Polling averages shift after late-cycle debate

22 min agoPennsylvania Michigan Arizona

Commodities190 Sources

Supply forecasts revised as export quotas tighten

1 day agoBrent OPEC Copper

Geopolitics430 Sources

Sanctions package reshapes trade-flow expectations

6 days agoEU Sanctions Tariffs

Licensed sources Structured context Global coverage Production-ready Agent-native Real-time

Backtest on the news as it existed then.

Your backtest is only as honest as its archive.

Future-data leakage

Thin scraped archives

Two different sources

A backtest you can actually defend.

One source for the backtest and the deployment.

Time-bounded retrieval

Historical News

News API

High-volume access

Knowledge Graph

MCP/CLI for Agents

400% more enrichments/extractions than competitors.

Chipmakers expand US capacity as AI infrastructure demand accelerates

Every source your research needs. One API.

Research that keeps going when others stop.

Reproducible from the first sweep.

From query to production in just three steps.

Set your as-of date

Sweep the archive

Deploy live

Most archives cannot tell you what was knowable when.

Built for systems that have to be right about time.

Find the right plan for you

Pay as you go

See details

Pro

See details

Spelunker

See details

Analyst

See details

Enterprise

See details

Validate on history you can actually trust.

We use cookies

Pay as you go

See details

Pro

See details

Spelunker

See details

Analyst

See details

Enterprise

See details