drillable
Ask anything. Every answer drills to its source. No guesses.
An honest handshake between your agent and you.
Ask anything. Every answer drills to its source. No guesses.
An honest handshake between your agent and you.
Mount it once over MCP. Your agent drills a cited source on these domains — and says nothing where there's nothing to hit.
Using drillable, what does it cover? — naming the tool in your first question teaches the agent to keep reaching for it.One question, two answers. A model answers from memory — just as confident when it's wrong. The grounded one cites the record, or hits a dry hole.
Under the hood, the grounded side is two parts: a model that does the talking, and a library of facts checked against named sources. The model doesn't answer from memory here — it has to look facts up.
Often both are right — a frontier model knows most facts. The point isn't who's right: one answer cites a source you can drill, flags when sources disagree, and will say “no record” rather than invent. That's what the checked library is for. Try to break it.
Every claim the verifier could not pin to an independent source — published per domain, sorted by why not. Most tools only show you what they have; this is the other table.
Plain words, because there isn't much to hide:
localStorage for your UI preferences.