17 algorithms · Free to start

The search API
that thinks.

One API call. 11 federated search backends, 8-signal neural ranking, 5-layer content extraction, and token-budgeted context with evidence graphs — in under 200ms.

POST /v1/search → 200 OK183ms
{
"query":"best async rust patterns",
"latency_ms":183,
"results":[ 11 sources × top 30 results ],
"ranked":[ HyperFusion 8-signal score ],
"content":"4 096 tokens extracted + cleaned",
"citations":8 (APA · IEEE · BibTeX),
"evidence":{ nodes: 14, edges: 22 }
}
11 backends · 30 results · 4 096 tokensAPI docs →
17
Novel algorithms
11
Search backends
930+
Tests passing
<200ms
Median latency

Why developers choose HyperSearchX

17 novel algorithms
not available anywhere else
< 200ms median
end-to-end search + extraction
11 search backends
federated in a single call
930+ unit tests
production-grade reliability
Built in Rust
zero GC pauses, safe concurrency
Zero telemetry
your queries stay private

The only search API with federated backends, neural ranking, and cross-session learning — features no competitor offers at any price.

See comparison
17 Novel Algorithms

Capabilities no other search API has

Every algorithm was designed from scratch to solve a real gap in the AI search stack — not a wrapper around an existing tool.

Unique

11-Backend Search Federation

Fan out a single query across SearXNG, Brave, GitHub, Reddit, HackerNews, StackOverflow, YouTube, Wikipedia, ArXiv, Bing, and DuckDuckGo — simultaneously. Adaptive Backend Selector picks the right mix per query intent. No other search API comes close.

Novel

HyperFusion Neural Ranking

8-signal ranking engine: BM25 lexical match + semantic similarity + temporal decay + domain authority + evidence density + source diversity + content depth + cross-source consensus. Results are measurably better than single-signal ranking — every time.

Novel

5-Layer CEP Content Extraction

Content Extraction Protocol: CSS selectors → Readability algorithm → Headless JS rendering → PDF parsing → Screenshot OCR. Every page — SPAs, paywalled content, PDFs — returns clean, structured text. Zero pages escape extraction.

Novel

QATBE Token Budget Control

Query-Aware Token-Budgeted Extraction scores every content segment with BM25 then solves a greedy knapsack to pack maximum relevance into your exact LLM context window. You always get the most useful content — not just the first N characters.

Novel

Deep Research Pipeline

AMRS multi-agent research swarm: 4 specialist agent types communicate over async channels, synthesise findings, and generate full evidence graphs with every claim traced to a source. Auto-generate citations in APA, IEEE, BibTeX, or Chicago.

Production

Production Resilience

Circuit breakers, bulkhead isolation, adaptive rate limiting, and latency prediction across all backends. Automatic failover, retry-and-refine with 5-checkpoint self-correction, and 99.9% uptime SLA on Pro and Enterprise plans.

Novel

YouTube & Social Intelligence

VideoFusion ranking for YouTube: transcript extraction, comment sentiment, teaching quality scoring. Native Reddit, HackerNews, and StackOverflow backends with community signal weighting. Social search no other API provides.

Novel

PIE Cross-Session Learning

Persistent Intelligence Engine tracks source trust, failure patterns, and query predictions across sessions via SQLite. HyperSearchX gets smarter with every query your application makes — improving result quality automatically over time.

Unique

MCP Protocol Native

First-class Model Context Protocol support. Expose all 17 algorithms as MCP tools to Claude, GPT-4, and any LLM with tool use. Structured, schema-validated outputs are AI-consumption-ready out of the box — no post-processing needed.

11+
Search backends
17
Novel algorithms
930+
Unit tests
< 200ms
Median latency
Pipeline Architecture

How it works

Six stages. Under 200ms. Every result traced back to its source with an evidence graph.

Swipe to see pipeline
Query
Federation
Ranking
Extraction
Token Budget
Response
01

Query Analysis

~2ms
QFD + QCE + QXE

Your query is fingerprinted, classified by intent, scored for complexity, and expanded with semantic variants. The system chooses the optimal backend mix before a single network call is made.

02

Multi-Backend Federation

~120ms
ABS + Resilience Layer

The Adaptive Backend Selector fans your query across up to 11 sources in parallel — SearXNG, Brave, GitHub, Reddit, StackOverflow, YouTube, and more. Circuit breakers handle backend failures invisibly.

03

HyperFusion Ranking

~18ms
8-Signal Neural Ranking

Results are scored on 8 signals: BM25 lexical match, semantic similarity, temporal freshness, domain authority, evidence density, source diversity, content depth, and cross-source consensus.

04

CEP Content Extraction

~40ms
5-Layer Cascade

Top-ranked URLs are deep-extracted via the Content Extraction Protocol: CSS selectors → Readability algorithm → Headless JS rendering → PDF parsing → Screenshot OCR. Zero pages escape clean extraction.

05

Token Budget Control

~5ms
QATBE Algorithm

Extracted content is segmented, BM25-scored for query relevance, then packed into your token budget via greedy knapsack. You always get the most relevant content that fits your LLM context window.

06

AI-Ready Response

Total < 200ms
Evidence Graph + Citations

The final response includes ranked results, extracted content within your budget, an evidence graph tracing every claim to a source, and auto-generated citations in APA, IEEE, BibTeX, or Chicago format.

Start building in 2 minutes
Get an API key, make your first search call, get back ranked results with extracted content and evidence graphs. Free plan — no credit card required.
Get API Key Free
Simple API

First result in 60 seconds

Install the SDK, paste your key, ship. Real multi-source search with zero boilerplate.

1import { HyperSearchX } from "@hypersearchx/sdk";
2
3const hsx = new HyperSearchX({
4 apiKey: process.env.HSX_API_KEY!,
5 baseUrl: "https://api.hypersearchx.zuhabul.com",
6});
7
8// Multi-source federated search
9const results = await hsx.search("rust async programming", {
10 backends: ["searxng", "brave", "github", "stackoverflow"],
11 maxResults: 10,
12 tier: "summary", // key_facts | summary | detailed | complete
13 tokenBudget: 2000, // QATBE greedy-knapsack packing
14 ranking: "hyperfusion" // 8-signal neural ranking
15});
16
17console.log(results.items[0].title);
18console.log(results.meta.tokensUsed); // always within budget
19console.log(results.evidenceGraph); // citations + trust scores
20
21// Deep-extract any URL
22const page = await hsx.extract("https://docs.rs/tokio", {
23 format: "markdown",
24 tokenBudget: 4096,
25 layer: "readability", // css | readability | headless | pdf | ocr
26});
$npm install @hypersearchx/sdk
API Response
200 OK187ms
Tokens
1984/2000
Backends
4
Ranked
10
Token budget usage99%
1Asynchronous Programming in Rust
brave

The async book covers futures, async/await syntax, Tokio runtime, and concurrent task management...

score
0.97
trust0.94
2tokio-rs/tokio — GitHub
github

A runtime for writing reliable, asynchronous, and slim applications. Stars: 28k...

score
0.93
trust0.98
3How does async/await work in Rust?
stackoverflow

Accepted answer (1.2k votes): Rust's async/await desugars into state machines at compile time...

score
0.88
trust0.91
Evidence Graph
10 nodes23 edgesconsensus 0.89
Honest Comparison

No other tool comes close

HyperSearchX is the only API combining search federation, neural ranking, deep content extraction, and cross-session AI learning — features competitors don't offer at any price point.

Scroll to compare
Feature
Best
HyperSearchX
Firecrawl
SerpAPI
Perplexity
Exa
Multi-source federation
11+ simultaneous backends
Token budget control (QATBE)
5-layer content extraction (CEP)
8-signal neural ranking
Evidence graphs + citations
Cross-session learning (PIE)
Deep research pipeline (AMRS)
YouTube & social search
Real-time monitoring + diffs
MCP protocol support
CLI tool included
Free tier included
Median API latency
Lower is better
< 200ms
~800ms
~400ms
~2000ms
~300ms
Full support
Partial support
Not available
87%
less cost vs Firecrawl Pro
faster than Perplexity

13 features that Firecrawl, SerpAPI, Perplexity, and Exa don't offer — even combined.

Transparent Pricing

Simple, transparent pricing

Start free. Upgrade when you need more. No surprise charges, no vendor lock-in, no hidden rate limits.

Free
$0forever

Evaluation, personal projects, and exploration.

1,000 req / month
60 req / min
  • All 11 search backends
  • 5-layer CEP extraction
  • HyperFusion ranking (8-signal)
  • Token budget management
  • Evidence graphs + citations
  • Community support (Discord)
Start Free
Starter
$19/ month

Indie developers shipping AI-powered products.

25,000 req / month
200 req / min
  • Everything in Free
  • 25,000 API requests / month
  • YouTube intelligence
  • Social media research
  • Real-time monitoring
  • Email support (48h SLA)
Start Starter
Most Popular
Pro
$79/ month

Teams and production AI applications.

250,000 req / month
500 req / min
  • Everything in Starter
  • 250,000 API requests / month
  • PIE cross-session learning
  • AMRS deep research pipeline
  • Priority support (4h SLA)
  • Usage analytics dashboard
Start Pro
Enterprise
Custom

Dedicated infrastructure, custom SLAs, full control.

Unlimited
2,000+ req / min
  • Everything in Pro
  • Unlimited API requests
  • Dedicated infrastructure
  • SLA guarantees (99.9% uptime)
  • SSO + team management
  • Dedicated Slack channel
Contact Sales

vs Firecrawl Pro ($599/mo): Our Pro plan ($79/mo) delivers a superset of features at 87% less cost. Federation, neural ranking, deep research, and cross-session learning that Firecrawl doesn't offer at any price.

All plans include all 17 algorithms, all 11 search backends, evidence graphs, and citations. No feature gating — only scale.