How do I make a request to the JUMP gateway?

POST to https://jump.adumu.com/api/jumpgateway/v1/execute/async with applicationSystemName, applicationNodeName and targetUrl. You get a jobUuid back immediately. Poll GET /api/jumpgateway/v1/job/{jobUuid}/status until the status reaches COMPLETED, then read scrapedData. You never hold a socket open waiting.

Which languages have a JUMP client library?

Eight: .NET, Node/TypeScript, Python, Go, Java, Rust, PHP and Perl. There is also a CLI binary, plus jump-curl and jump-wget as drop-in replacements that take the same flags.

Does JUMP parse the page for me?

No, and that is deliberate. The response body comes back raw and unmodified. Extraction stays with the caller, so a change to your parsing never requires a change to your gateway plan.

How long are results kept?

Response bodies stay readable for 24 hours. The record of the run, minus the payload, is kept for 90 days.

Can an AI agent call JUMP directly?

Yes. JUMP ships a Model Context Protocol server exposing the same submit and poll contract under the same field names, so an agent gets the raw body back and parses it itself.

Web Scraping API Documentation & SDKs

Data Flow & Architecture Comparison

How It Works: In-House Scrapers vs. JUMP Gateway

Step 1 · Client

1 Single API Request

Your script, service, or Claude Desktop agent fires POST /api/jumpgateway/v1/execute/async or the MCP tool.

Input: 6 clean fields · Zero browser setup

Step 2 · Gateway

3-Layer Stealth Engine

Layer 1: TLS 1.3 JA4 stealth · Layer 2: In-memory headless sandbox · Layer 3: Pristine IP mesh.

Clearance: Automated JS & challenge solve

Step 3 · Target

Protected Web Targets

Cloudflare, Akamai, Datadome, and login walls see an authentic human browser session. Zero 403 blocks.

Wire: 100% 200 OK unmetered egress

Step 4 · Ingest

Clean Structured Output

Parsed JSON streams directly into PostgreSQL, AWS S3, BigQuery, or your AI Agent context window.

Delivery: Real-time webhook or DB write

Success Rate 99.8% Cleared

Execution Latency 1.2s – 3.4s Avg

Pricing Model From $99 / mo Flat

Scraper Maintenance 0 hrs / week

• JUMP SDK · THE COMPLETE CLIENT SCRIPT SUBMIT · POLL · READ

// Submit — returns a job id, no socket held open
const { jobUuid } = await jump.executeAsync({
  applicationSystemName: "PRICE_WATCH",
  applicationNodeName: "NODE_01",
  targetUrl: "https://example.com/catalogue?page=2"
});

// Poll — the body comes back raw. Parsing stays yours.
const result = await jump.waitFor(jobUuid);

Step 1 · Client

Fragile Scraper Script

45+ lines of Puppeteer boilerplate, manual proxy rotation logic, user-agent arrays, and retry loops.

Failure: Breaks whenever site changes DOM

Step 2 · Proxies

Proxy Bandwidth Burn

Third-party proxy rotators deliver burned IPs. Frequent 429 rate limits and thousands in bandwidth charges.

Cost: $2,500 – $6,000 / mo proxy fees

Step 3 · Headless

RAM Leaks & WAF Blocks

Chrome nodes crash server memory. Cloudflare & Akamai flag TLS 1.3 JA4 fingerprints with 403 Forbidden.

Alert: 403 Forbidden · CAPTCHA loops

Step 4 · Ingest

Corrupted Pipelines

Incomplete data streams, missed price changes, stalled AI agent runs, and emergency engineering fire drills.

Overhead: 18+ hrs / week maintenance

Block / Error Rate 35% – 45% Blocked

Avg Pipeline Latency 8.4s + Retries

Metered Competitor Cost ~$3,750 / mo (500k pgs)

Engineering Burden 18 hrs / week fixing

• IN-HOUSE PUPPETEER · 45+ LINES OF FRAGILE BOILERPLATE HIGH FAILURE RATE

// Fragile In-House Scraping Stack with Proxy Rotation & Anti-Detect Hacks
const browser = await puppeteer.launch({
  args: ['--no-sandbox', '--proxy-server=http://proxy-provider:port', '--disable-blink-features=AutomationControlled']
});
const page = await browser.newPage();
await page.authenticate({ username: '...', password: '...' });
try {
  await page.goto(targetUrl, { waitUntil: 'networkidle2', timeout: 30000 });
  // ⚠️ Error 403: Cloudflare Challenge detected (JA4 TLS mismatch)...
} catch (err) { /* Retry loop with next proxy... */ }

in-house headache

TLS & Browser Leaks

Modern defenses inspect TLS 1.3 frame signatures (JA4) and headless flags before running your script. Patching them in-house is an endless engineering sink.

proxy costs

Proxy Pool Burn

Managing residential proxy pools costs thousands per month and bad IPs burn quickly. JUMP manages pristine dedicated addresses transparently.

the JUMP answer

Zero-Fluff Gateway

Submit a target URL via SDK or MCP tool. Receive clean structured JSON. Concurrency is dedicated and unmetered.

Client SDKs & Agent Tool Schemas

Drop-in implementations across 7 native language SDKs, CLI tools, and Model Context Protocol (MCP) servers.

• GATEWAY SPEC

// Model Context Protocol tool definition — Claude Desktop / Cursor / agents
{
  "name": "jump_execute_async",
  "description": "Submit a page to the JUMP gateway. Returns a jobUuid to poll; the raw body comes back on completion.",
  "parameters": {
    "type": "object",
    "required": ["applicationSystemName", "targetUrl"],
    "properties": {
      "applicationSystemName": { "type": "string", "description": "Which of your systems is calling" },
      "applicationNodeName":   { "type": "string", "description": "Which node of it — labels you choose" },
      "targetUrl":             { "type": "string", "description": "The page to fetch" }
    }
  }
}

Ecosystem Registry Package Identifier Import / Usage

.NET (C#) NuGet Adumu.Jump.SDK using Adumu.Jump;

TypeScript / Node npm @adumu/jump import { JumpClient } from '@adumu/jump'

Python PyPI adumu-jump import adumu_jump

Java Maven Central com.adumu:jump-sdk import com.adumu.jump.JumpClient;

Go Go modules github.com/adumu/jump-go import jump "github.com/adumu/jump-go"

Rust crates.io adumu-jump use adumu_jump::Client;

PHP Packagist adumu/jump-sdk use Adumu\Jump\Client;

Perl CPAN Adumu::Jump use Adumu::Jump;

Claude / Cursor Anthropic MCP jump-mcp-server mcpServers: { "jump": { ... } }

Anything else GitHub release One binary jump-curl / jump-wget — same flags, one word changed

Ecosystem	Registry	Package Identifier	Import / Usage
.NET (C#)	NuGet	`Adumu.Jump.SDK`	`using Adumu.Jump;`
TypeScript / Node	npm	`@adumu/jump`	`import { JumpClient } from '@adumu/jump'`
Python	PyPI	`adumu-jump`	`import adumu_jump`
Java	Maven Central	`com.adumu:jump-sdk`	`import com.adumu.jump.JumpClient;`
Go	Go modules	`github.com/adumu/jump-go`	`import jump "github.com/adumu/jump-go"`
Rust	crates.io	`adumu-jump`	`use adumu_jump::Client;`
PHP	Packagist	`adumu/jump-sdk`	`use Adumu\Jump\Client;`
Perl	CPAN	`Adumu::Jump`	`use Adumu::Jump;`
Claude / Cursor	Anthropic MCP	`jump-mcp-server`	`mcpServers: { "jump": { ... } }`
Anything else	GitHub release	One binary	`jump-curl` / `jump-wget` — same flags, one word changed

ℹ️

SDK Availability & Specification Notice: Code examples and method signatures reflect the unified Adumu JUMP API specification and may undergo minor refinements prior to final release. Select native language client packages and public registry distributions are currently in staged private preview and will be published across public package registries (NuGet, npm, PyPI, crates.io, Maven Central, Packagist) in our upcoming platform release. All gateway features and concurrency lanes remain 100% operational today via direct HTTP / cURL and Model Context Protocol (MCP).

The core request contract

Six fields define a complete enterprise scrape job. Everything else has a sane default.

Field Type Description

applicationSystemName string (required) Which of your systems is calling. A label you choose — it groups runs in the console.

applicationNodeName string (required) Which node of that system. Also a label you choose.

targetUrl string (required) The page to fetch. Login walls, challenge walls and SPAs are handled by the gateway.

engine string (optional) Pin an engine. Left alone, the gateway matches one to the target.

commands array (optional) Click, type, wait, scroll — run in the order listed. Empty means the page is waited for and read.

snapshots array (optional) Label a capture and it returns as its own entry, so one run can hand back several moments.

walk object (optional) A list of URLs, the pages after a starting one, or links followed from a seed — each with its own per-page rules.

Field	Type	Description
`applicationSystemName`	string (required)	Which of your systems is calling. A label you choose — it groups runs in the console.
`applicationNodeName`	string (required)	Which node of that system. Also a label you choose.
`targetUrl`	string (required)	The page to fetch. Login walls, challenge walls and SPAs are handled by the gateway.
`engine`	string (optional)	Pin an engine. Left alone, the gateway matches one to the target.
`commands`	array (optional)	Click, type, wait, scroll — run in the order listed. Empty means the page is waited for and read.
`snapshots`	array (optional)	Label a capture and it returns as its own entry, so one run can hand back several moments.
`walk`	object (optional)	A list of URLs, the pages after a starting one, or links followed from a seed — each with its own per-page rules.

Build with JUMP. From $99 a month.

Start with a dedicated cloud lane from $99 a month flat. Complete API key, SDK access, and console link delivered in 24 hours.

Start a lane — $99 / mo View full rate card

One endpoint. Any target. Six fields.