Plug your AI into the live web

Web data,
ready for AI.

Turn any URL into clean markdown. Search the web by keyword. Or plug your AI assistant straight into the live web with our MCP server.

app.chuger.com/dashboard

Dashboard

Last 7 days

Credits used

24,103+12%

Requests

12,401+8%

Avg latency

187ms-3%

Requests over time14d

Recent requests

GET/v1/content200
GET/v1/serp200
POST/v1/content/bulk200
GET/v1/content200
GET/v1/scrape200

The problem

AI needs the web. The web isn't AI-ready.

Three small pains stand between your AI and useful web data. Chuger collapses them into one API.

HTML isn't markdown

Pages come with nav, footers, ads, scripts. Your LLM doesn't need any of it. Cleaning it yourself is a project, not a feature.

Search and scrape live in different APIs

Find URLs with one tool. Scrape them with another. Two keys, two bills, two retry layers — and the glue is on you.

AI assistants can't browse on their own

Claude, Cursor, and other clients want tools, not REST. Most scraping APIs don't speak MCP. Chuger does, out of the box.

Three ways to use Chuger

Three integrations. One credit pool.

Pick the integration that fits how you build. They share a token, an envelope, and a bill.

For AI assistants

Plug your AI in directly

Drop our MCP server into Claude Desktop, Cursor, or any MCP-aware client. Your assistant can scrape, extract, and search the live web — no glue code.

One config line
Works in any MCP client
Same credits as the API

For content extraction

Any URL → clean markdown

Call /content with a URL. Get back markdown — title, body, author, meta — already cleaned of nav, ads, and boilerplate. Use /scrape if you want raw HTML.

Markdown or HTML
Bulk up to 500 URLs
Webhook delivery for jobs

For discovery

Find pages by keyword

Call /serp with a query and get search results. Or /serp/content to search and extract clean content from every result in a single trip.

Search any region
Combined search + extract
Positions and snippets

Tools

Six tools. One API.

Same auth, same envelope, same error model on every call. Credit costs are published — no surprise overage bills.

GET/v1/scrape

Scrape

Raw HTML for any URL. Pass a target, get the page.

1 creditJSON · streaming where applicable

GET/v1/content

Content

URL → clean markdown with title, body, and metadata. LLM-ready.

2 creditsJSON · streaming where applicable

POST/v1/content/bulk

Bulk Content

Up to 500 URLs per call. Webhook delivery — no queues to manage.

2 credits eachJSON · streaming where applicable

GET/v1/map

Map

Discover every URL a site publishes via its sitemaps. Map first, extract only what you need.

3 creditsJSON · streaming where applicable

GET/v1/serp

SERP

Search any keyword. Positions, snippets, sitelinks — one envelope, every region.

5 creditsJSON · streaming where applicable

GET/v1/serp/content

SERP + Content

Search and extract clean content from each result in a single call.

15 creditsJSON · streaming where applicable

POST/v1/prompt

Prompt

Send a question, optionally grounded in page content for 2 additional credits.

10 creditsJSON · streaming where applicable

Code

Drop in. Ship today.

Pick your language. The shape is identical across tools — once you've integrated one, you've integrated all six.

# Any URL → clean markdown for your LLM
curl "https://api.chuger.com/v1/content?url=https://stripe.com/docs/api" \
  -H "Authorization: Bearer $CHUGER_KEY"

# → 200 OK · 142ms · cost=2 credits
# → { "title": "...", "markdown": "...", "meta": {...} }

What's in the box

The boring parts, productized.

MCP server, included

A first-class Model Context Protocol server ships with every plan. Wire Chuger into Claude Desktop, Cursor, or any MCP-aware client in one config line — or paste our agent install prompt and let your AI set itself up. Guide at docs.chuger.com/mcp.

Markdown-ready extraction

Skip the cheerio gymnastics. /content returns structured markdown with title, body, author, and meta — designed for direct ingestion into LLMs.

Prompt the page

Send a URL and a question to /prompt — get an LLM answer grounded in the live page. Skip the fetch-then-prompt boilerplate.

Built-in fallback

Every request runs through a live health-aware router. When a path degrades, traffic re-routes before you see a failure. We never expose the underlying chain.

Built for

Workflows that need fresh, structured web data.

AI agent data ingestion

RAG pipeline content sourcing

Real-time research in chat apps

SEO and SERP rank monitoring

Lead enrichment from public profiles

Content aggregation and summarization

Reliability

Reliability, productized.

Health-aware routing isn't a feature flag — it's how every request runs. Health checks, real-time error rates, and instant rerouting are the default.

99.99%

Routing uptime

Aggregate across all tools, last 90 days. Live numbers on the status page.

<200ms

Reroute overhead

p95 latency added when a primary path degrades and traffic re-routes mid-flight.

Failed-request cost

If we don't return a 2xx, you don't pay. Including upstream failures and timeouts.

Pricing

One credit pool. Every tool.

Choose a fixed monthly allowance. Failed requests do not consume credits, and there are no automatic top-ups.

View current plans

FAQ

Questions, answered.

Call /v1/content with a URL. You get back clean markdown — title, body, author, meta — ready to drop into an LLM context window. Use /v1/scrape if you want raw HTML instead.

Feed the web to your AI.

Get clean web data into your AI in minutes. Free to start, pay as you grow.