Finlight delivers the same enriched financial news article over REST, WebSocket, webhooks, and an MCP server, and Kevin Bartsch’s REST vs WebSocket vs Webhooks vs MCP: Which to Use is the company’s guide to picking one. The framing is refreshingly unpretentious: “They aren’t competitors; they’re tools for different jobs.” REST is a pull model for search, historical backfills, and dashboards that refresh on a schedule. WebSocket “sends each matching article the moment it’s published, with no polling,” in enhanced and raw variants. Webhooks call your HTTPS endpoint the moment an article matches your filters, signed so you can verify the sender, also in enriched and raw variants. The MCP server connects Finlight to Claude or ChatGPT so you can ask for market news in plain language. The closing advice is the useful part: “REST for the initial backfill, then WebSocket or webhooks for everything live.”
There are no vendor figures to scrutinize here, which is unusual for a provider post and to Finlight’s credit. The one claim worth testing is structural: that four delivery methods exist and carry the same data. The catalog record describes Finlight as a real-time news API with “sub-second latency streaming,” and the provider’s own tags on apis.io include both Webhook and WebSocket. So the claim that the streams exist checks out against what the catalog has already recorded. What the catalog cannot confirm is how they are shaped, and that turns out to be the story.
The Finlight provider page lists two API pages, and both are the REST half of the story. The Finlight Articles API is the search and backfill surface the post tells you to start with, and the Finlight Sources API is the list of publishers behind it. The MCP server dimension on the Agent Readiness score is lit, so the fourth delivery method is confirmed at the catalog level. The agentic access profile maps 3 operations, 1 of them acting, which is what a news API should look like: mostly reads, one place where an agent changes state.
The Kin Score is 40.3, developing band, and the Agent Readiness score is 39.2. Discoverability carries it at 68.5, access clarity at 63.2, contract quality at 59.5. Operational transparency is 0.0 and contract governance is 9.8. The dimension that matters for this post is event surface described, and it is unlit. Finlight ships two WebSocket variants and two webhook variants with signed deliveries, and none of that is described in a machine-readable contract the catalog can read. Half of what the guide explains exists only in prose. A provider that took the trouble to build four doors has published the blueprint for two.