search & execution
engine built for
ai agents.
give your agent one api to search sources, act on pages, verify claims, and return answers with evidence.
product
one api. live sources, browser state, proof.
nex combines retrieval, execution, and verification in one system so agents can return answers you can inspect.
ranked evidence with citation metadata
stateful runs across real pages
verification mapped back to evidence
answers with citations and trace data
primitives
what your agent can actually do.
compact runtime primitives that describe the actual system surface.
source ranking
rank live sources by task relevance and return citation-ready evidence.
browser state
inspect page state, interact with forms, and keep the run replayable.
structured extraction
turn pages, docs, and app states into schema-shaped data.
claim verification
check claims against evidence and surface weak reasoning before output.
built as an api first.
plug nex into an agent and give it one interface for live retrieval, browser tasks, extraction, and verification.
const result = await nex.run({
task: "compare browser agent startups and return a verified market map",
mode: "retrieve+execute+verify",
retrieval: {
sources: ["web", "docs"],
citations: true
},
execution: {
browser: true,
replay: true
},
verification: {
claims: true,
confidence: true
},
output: {
format: "json",
includeTrace: true
}
})
console.log(result.answer)
console.log(result.sources)
console.log(result.trace)difference
not just search. not just browser automation.
search stops at retrieval. browser tools stop at interaction. nex connects both and returns evidence.
agentic tasks
how research, ops, and coding agents are used.
simple runtime examples that show the system doing real work.
01
research agent
build market maps and compare companies from live sources.
sourced summary with replayable proof
02
ops agent
collect contacts, validate vendors, and complete live workflows.
verified records ready for ops
03
coding agent
search current docs, inspect implementations, and verify api usage.
implementation context with citations
architecture
a simple runtime path.
agent tasks move through four steps before nex returns the result.
retrieve
sources are ranked, parsed, and kept citation-ready.
execute
browser state is inspected and actions are captured.
verify
claims are checked against evidence before output.
answer + trace
the agent receives a sourced answer and replayable proof.
get early access to nex.
we are onboarding teams building agents that depend on live retrieval, browser execution, and verifiable outputs.
prefer email? amirlan@extensy.dev
faq
straight answers.
what is nex?
nex is an api for agents that combines retrieval, browser execution, verification, and source-backed output.
how is this different from search apis?
search apis stop at links and snippets. nex also executes tasks, checks claims, and returns evidence.
how is this different from browser automation?
browser automation clicks and types. nex adds retrieval, extraction, verification, and citations around those actions.
does nex use real browsers?
yes. browser execution is part of the runtime when a task needs page state, screenshots, forms, or live interaction.
what are replayable traces?
traces are structured records of sources, browser actions, evidence, and verification steps.
does nex return citations?
yes. answers can point back to the sources that support them.
who is this for?
technical teams building agents that depend on live sources, browser state, and verifiable output.
when will api access open?
we are onboarding early teams now and prioritizing builders with production tasks that require retrieval, execution, and verification.