throughline

Find the news that matters to the markets you follow.

Throughline is a free, rate-limited API that tells you which news articles are relevant to which prediction-market contracts, and why. Instead of reading every headline and working out for yourself what bears on a contract, you get the article already matched to the contract, with the reason stated. Geopolitical markets first.

one article, three contracts matched illustrative example, not live output

reuters · 2026-11-03T08:14Z · en

Israel and Lebanon agree a framework to demarcate the land border

resolves to

  • Israel-Lebanon border agreement signed before 30 June 2027?

    The framework names a demarcation mechanism and a working timetable, which is the step a signed agreement in the window depends on.

    assessednewimpact 3evidence: full text

  • Does Hezbollah begin verified disarmament in 2027?

    Border demarcation has been a stated precondition in previous disarmament proposals. The link is plausible and not yet examined closely.

    preliminaryupdateimpact 2evidence: headline only

  • Israel-Syria normalization announced in 2027?

    Withdrawn on closer reading. The article does not mention Syria; the match came from a shared regional entity and does not survive.

    withdrawnwas: preliminaryrevision 2

What you get for each match

A match says one article is relevant to one contract. It is not a forecast, and the fields are chosen so you can judge the match rather than trust it.

article
Publisher, link, language and the original publication time, so you can go read it.
contract
The exact market and rule revision matched, not a market title that may have been edited since.
reason
Why this article was matched to this contract, in words, traceable to the evidence it was drawn from.
novelty
new, update, repeat or uncertain — so the fifth retelling of a development does not read like a fifth development.
impact
0 to 4, a model's reading of how much the story bears on the contract. Not a probability and not calibrated confidence.
evidence
What the assessment actually saw: the full text, a summary, or only a headline. Labeled, because it changes how much the reason is worth.
uncertainty
Stated plainly, including the case where the match itself is uncertain rather than the outcome.
clocks
Four separate times: when the source published, when it was received, when it was assessed, and when it became publicly readable. They are not the same number and are never collapsed into one.
identity
A stable match id and a revision, so a correction is a new revision of the same match rather than a second, contradictory match.

Matches get corrected, and you see it happen

Matches publish as soon as they exist, which means most of them arrive before anything has looked at them carefully. That earliness is the point, so the stage is published alongside the match instead of being hidden behind it.

preliminary
A first-pass match. Fast, broad, and the stage most matches are in when you first see them.
assessed
A closer reading has been done: summary, per-contract reasoning, impact and missing context.
withdrawn
The match did not survive a closer reading. It stays readable at its final revision rather than disappearing from your copy.

A stage is not a truth certification. An assessed match is one that was examined, not one that was verified. Not every preliminary match receives a second look, so an old preliminary match means nobody got to it — it does not mean the match was confirmed by silence.

Telling a quiet market from a broken pipeline

A quiet feed is ambiguous, and most data products resolve that ambiguity in their own favour by returning an empty list. There is a separate endpoint that distinguishes the four states, because a market's last match is not its last screen.

What it won't publish

Throughline is built and operated alongside a system that trades these markets. That makes the boundary between the two the most important thing on this page, so it is drawn structurally rather than promised.

Published: whether a story bears on a contract, and why. Every one of those fields is price-independent — it reads the same whatever the contract is trading at.

Never published: forecasts, probabilities, positions, financial eligibility, and anything that answers is there edge against the current price. That last question is the position itself, and the component that answers it, analyst_screener, has no path to the public feed.

The published set is not narrowed by any trading predicate either — no volume floor, no price band, no liquidity filter. Relevance is published within the declared coverage universe regardless of our position or current edge, because a feed shaped by a trading predicate is a map of the operator's book no matter how clean its fields are.

Not yet serving

The analysis flow behind this runs today. The public feed and the API do not exist yet — the publication contract is still being written, and nothing above should be read as a description of a running service. Access will be free and rate-limited when it opens, with optional email updates for contracts you follow.

If this would be useful to you, the most valuable thing you can send is which contracts you would want covered and what would make the matches worth reading.