Atoms / Library

featureperformancewarmingqualifyABMMCPcold-cache

Warm cache hits land in about 50ms even on a first visit

10 to 20 seconds

Previous cold build wait

~50ms

First click latency after warming

0 (same 1 personalisation charge)

Extra cost vs unwarmed mint

Spectare's cold-build path used to mean a 10 to 20 second wait before the largest paint on a fresh intent. Mint-time warming closes that gap. When a key-authenticated qualify call or a qualify_visitor MCP tool call produces a link, Spectare now builds the classified content cell before the mint returns, so the first click on that link is served from cache in around 50ms rather than triggering a live Sonnet build.

The cost accounting is honest: the warm build is metered as one personalisation, exactly the same charge the click would have incurred anyway. The work is moved earlier, not added. If you pass warm: false in the mint request, warming is skipped and the link lands cold as before. Confidence-zero intents are also skipped automatically because their serve path uses default atoms and needs no pre-built cell.

The mint response now reports warmed: true when a cell was pre-built, so your tooling can confirm the guarantee was applied before the link goes into an inbox or an agent message. Anonymous page traffic is unaffected because its own assemble call follows immediately on arrival and warming there is already handled by construction.

This is the right trade for ABM outreach and agent-generated links: a slightly slower mint that gives every recipient a sub-100ms first paint is worth more than a fast mint that makes every recipient wait.

This is one atom from the Spectare content library. Spectare assembles the right atoms for each visitor in real time, based on who they are and how they arrived.