About unlob

Why this exists

Every search API in production was designed for a human reader and adapted for agents afterwards. The adaptation is usually cosmetic — cleaner JSON, a summary field — while the underlying assumption stays in place: return a lot, rank it well, and let the reader discard the rest.

That assumption is free for a person and expensive for an agent. Every irrelevant result costs context and reasoning tokens, and the agent cannot ask a clarifying question or tell a good source from a plausible one without help. So we built the index around a different rule: admit only what is worth returning, and give the agent the structure it would otherwise rebuild in its context window.

What we actually built

An index that is bounded on purpose rather than endlessly growing, because for a machine reader ten mediocre results are a bill rather than a convenience. A coverage graph published from signals the index already computed, so an agent can traverse relationships instead of reconstructing them in its context window. And a removal ledger, so when something is not in the index we can say why rather than returning silence.

The economics follow from the architecture rather than from a pricing decision. We own the index, so no part of our bill is a per-query fee to an upstream provider, and what we do spend tracks how much the API is used rather than how large the index has grown. We are self-funded and not raising, so there is no valuation the price has to grow into. Thecost model sets out the argument in full — including what we no longer publish about how the index is built, and why.

What we are not

Not the biggest index. Exa and Parallel are valued at $2.2 billion and $2.0 billion, both run their own index, and both have more of the web than we do. Brave has 40 billion pages. If raw coverage is your binding constraint, one of them is probably the better choice, and ourcomparison pages say so specifically rather than pretending otherwise.

Not venture funded, and not raising. That means our pricing has no valuation to grow into. It also means we are smaller than everyone we compete with, and you should weigh both halves of that.

Not open source. The engine is a closed commercial product. What we publish is the architecture, the economics and the measurement conditions behind every number — enough to evaluate the claims without shipping the implementation.

How we operate

Every benchmark on this site carries the conditions it was measured under. Every competitor figure carries a link to the vendor's own page and the date we read it. Every comparison page names at least one scenario where the competitor is the better choice. Every migration guide has a section listing what you will lose.

That is not modesty, it is self-interest: in a market where every vendor publishes a comparison table that happens to favour them, being checkable is the only durable differentiator in the content itself.

Who

unlob is built by Dipankar Sarkar. Questions, removal requests and coverage complaints all reach a person athello@unlob.com.