INTELLIGENCE, WITH GRAVITY

Everything
relevant,
in orbit.

Orbita helps your AI find the ideas, facts and pages that pull their weight — without getting lost in the universe of the web.

Enter orbit →How it works ↓
01 / HIGH GRAVITYA systems view of carbon removalJournal of Climate
02 / HIGH GRAVITYFoundations for durable climate actionnature.org
50 sources in orbit

SEARCH WITH A SENSE OF WONDER

Context doesn’t live
on one page.

When answers are scattered, an agent needs more than a search box. Orbita understands the relationships around every finding, then draws the useful ones close.

Meet the Orbita API ↗

01

Find the
center.

Intent-aware retrieval finds the source that starts the real answer.

02

Trace the
connections.

Relevance expands through related pages, people, and ideas.

03

Keep it
fresh.

A living web index stays in step with what changed today.

ORBITA GRAVITY / EVIDENCE QUALITY

Noise drifts out.
Evidence stays.

Gravity gives agents cleaner, verifiable source material without publishing Orbita’s internal recipe.

See measured outcomes ↗

LESS NOISE

page chromeirrelevant repetitionlow-value clutter

USEFUL OUTPUT

verifiable detailsuseful source materialcitation-ready context
71.3%less irrelevant page text
83.5%exact fields retained from pages
93.5%exact fields across supported sources
91.0%exact evidence in a bounded response

Frozen test: five public admissions sites and ten pages. These are Orbita before/after measurements, not a competitor benchmark.

PUBLIC NUMBERS FROM OTHER SERVICES

Their published prices, limits and vendor-run measurements—without exposing Orbita’s implementation.

Exa

$1 per 1k pages for each content type. Vendor 250-URL eval: 89.3 accuracy, 96.7 code recall.

Official source ↗
Tavily

Up to 20 URLs per call and 1–5 focused chunks up to 500 characters each.

Official source ↗
Linkup

$1 per 1k pages without JavaScript; $5 per 1k with JavaScript. One URL per call.

Official source ↗
Firecrawl

One credit per page. Vendor 1,000-URL eval: 96% success and 3.387 s p95.

Official source ↗

YOUR NEW FAVORITE ENDPOINT

A little call.
A lot of universe.

Ask once. Receive ranked results, rich content and clean citations — all with relevance you can explain.

Try the API →
ORBITA / QUERY
orbita.search({
  query: "biomimicry in architecture",
  mode: "standard",
  citations: true
})

STANDARD → 100 candidates → up to 10 results

MEASURED AND ESTIMATED SEPARATELY

Latency without
fantasy.

Orbita is not deployed in production yet, so we do not claim user latency. The map shows only the theoretical round-trip floor through fiber from a planned Lviv region; real routes will be slower.

≥3.4ms · theoretical fiber RTT floor
339 km
great-circle distance from planned Lviv
91–98 ms
controlled local diagnostic
not measured
production end-to-end

Controlled local diagnostic only. It excludes production networking and service overhead and is not an architecture disclosure.

Choose a city →

FROZEN OUTPUT REPLAY · VENDOR CLAIMS SEPARATE

One test.
No made-up numbers.

100 frozen known-page searches: 22 factual and 78 title-navigation targets across 26 hosts. The original 100-query responses are unchanged. Orbita’s current Fast-to-Standard consistency rule was replayed over those frozen outputs; Standard now leads without a manual score adjustment.

      Rank metrics are a deterministic replay of the unchanged 2026-09-12 responses under the corrected mode contract—no new calls and no manual score changes. Latency remains the original historical measurement, not a post-correction SLA.

      PUBLISHED BY THE PROVIDERS · NOT OUR TEST

      Useful context from current official pages. These claims are labelled and never mixed into the measured chart.

      Exa<350 ms p50

      Vendor-run Fast benchmark; public price $7/1k searches up to 10 results.

      Official source ↗
      TavilyNo public numeric SLA

      Basic/Fast is $5–8/1k at published plan and PAYG credit prices.

      Official source ↗
      Linkup1–3 s

      Vendor range for Standard; raw results are $5/1k and the homepage claims 99.9% SLA-backed uptime.

      Official source ↗
      Serper1–2 s

      Vendor says “usually”; prepaid price is $0.30–1.00/1k. Not in our run: this project has no current Serper key.

      Official source ↗

      FROM PILOT TO PRODUCTION

      Connect quickly.
      Grow on your terms.

      One API to start. As your team and traffic grow, we align limits, support, data location and SLA without forcing a new integration.

      Pilot
      Test Orbita on your own queries without a long implementation cycle.
      Control
      Set the budget, access, retention and usage boundaries your product needs.
      Scale
      Move to higher limits and agreed service terms after a successful pilot.

      FIVE FIXED DEPTHS + AUTO

      Five depths.
      Plus Auto.

      Pay for the reach the question actually needs. Every mode returns ranked sources and citation-ready passages; publication dates are included when the source exposes one.

      Start free ↗

      01

      Fast

      20candidates
      →
      5results

      Autocomplete, quick facts and routing.

      —New Search · per 1 000
      —Search + Compact
      max — context tokens

      02

      Standard

      100candidates
      →
      10results

      Daily agents, chat and product search.

      —New Search · per 1 000
      —Search + Compact
      max — context tokens

      03

      Pro

      200candidates
      →
      25results

      Comparisons and evidence-heavy questions.

      —New Search · per 1 000
      —Search + Compact
      max — context tokens

      04

      Deep

      300candidates
      →
      50results

      Research across many independent sources.

      —New Search · per 1 000
      —Search + Compact
      max — context tokens

      05

      Ultra

      500candidates
      →
      100results

      Maximum recall for critical investigations.

      —New Search · per 1 000
      —Search + Compact
      max — context tokens

      06

      Auto

      Autorequested
      →
      Fast–Ultraresolved mode

      Orbita chooses the fixed depth that fits the question and your limits.

      —resolved Search price
      —resolved Compact price
      Never exceeds max_mode, max_cost_usd or max_results

      Five explicit depths plus Auto, which resolves to one of them and uses its price. Compact adds a controlled context pass.

      • New Search—
      • Search + Compact—
      • Linkup$5
      • Tavily$5–8
      • Exa$7
      Published rates checked 2026-09-12. Products and output depth differ.

      OUR PRINCIPLE

      The web can be endless.
      Search should not feel chaotic.

      Orbita pulls the right sources closer, so AI works with evidence instead of getting lost among pages.