Newstools for publishers
Source discovery and monitoring
Spot new stories and relevant signals before they get lost in the noise.
Let's talk about your project
A shortlist, not a flood
Deduplication, relevance, and clustering by topic before anyone has to look.
The newsroom decides
Editors choose in the panel which candidates move on to research or drafting.
Sources and costs under control
Health monitoring for every source, and cost controls when the analysis uses AI models.
From hundreds of sources to a shortlist
A person decides at this step
- Configured sourcesRSS feeds, sitemaps, and your own sources, with categories, thresholds, and schedule defined with the newsroom.
- Scheduled captureThe system scans the sources at the set frequency and records the health of each one.
- FilteringAgents deduplicate, classify by topic, and analyze relevance against the agreed thresholds.
- ClusteringPieces that cover the same story are grouped with embeddings, and references are extracted.
- Editorial decisionIn the panel, editors choose which candidates move on to research or drafting and discard the rest.
- ExportThe chosen candidates go out to the configured destination, with their references.
The agenda is still decided by the newsroom.
What it solves
New stories surface across hundreds of sources at once: RSS feeds, other titles' sitemaps, and any source you configure. Following them by hand eats hours at the news desk, and what matters gets lost among repeats and noise.
We build systems that scan the sources you define, deduplicate, classify by topic, analyze relevance, and group the pieces that cover the same story. The result is a shortlist of candidates, not a flood of links.
How we work
We define sources, categories, relevance thresholds, and export targets with the newsroom. Capture is scheduled: each project sets its own frequency and rules.
On what is captured, agents deduplicate articles, classify topics, analyze relevance, and group related stories with embeddings. They can extract references and export findings for review. Editors see the candidates in a panel and decide which ones move on to research or drafting.
When the analysis uses AI models, we include cost controls from day one, and we monitor the health of every source to know when it stops responding or changes format.
What we deliver
- A capture system configured with your sources, categories, schedule, and destinations.
- Deduplication, classification, relevance analysis, and semantic clustering of related stories.
- A review panel where the newsroom tracks the work, source health, and the candidates awaiting a decision.
- Export of candidates and references to the destination you choose: research, draft generation, or your own workflow.
What comes out of the system are proposals with context: each candidate arrives with its topic, its relevance analysis, and the related pieces already grouped, so deciding takes less time than searching. The agenda is still decided by the newsroom.
Who it is for
News desks and breaking-news teams that monitor many sources, specialized sections that follow very specific sources, and publishing groups that want shared monitoring for several titles without each one repeating the work.
What we have already built
Discover Sources is the discovery module we start every deployment from: it scans RSS feeds and sitemaps, deduplicates, classifies topics, analyzes relevance, and groups related pieces with embeddings. Each project defines its sources, categories, schedule, and destinations, and its dashboard tracks the work and source health. We adapt it to your newsroom instead of building it from scratch.