Live in Production · mcp.datamint.xyz

Model Context Protocol (MCP)

Equip Cursor, Claude Desktop, Claude Code, Windsurf, or custom LangChain and Python agents with native web scraping and extraction capabilities in under 30 seconds.

Endpoint: https://mcp.datamint.xyz/sse

Transport: SSE (Server-Sent Events) & Stdio

.cursor/mcp.jsonUTF-8

{
  "mcpServers": {
    "datamint": {
      "url": "https://mcp.datamint.xyz/sse"
    }
  }
}

Tip: Add to your project's .cursor/mcp.json or global Cursor Settings → Features → MCP Servers.

Quickstart: Connect in 30 Seconds

The datamint MCP server is hosted at https://mcp.datamint.xyz/sse. Any MCP-compatible client can connect via Server-Sent Events (SSE) without cloning code, managing Node.js processes, or building containers.

When your agent connects to the hosted endpoint, it immediately acquires tools to mint web pages, extract structured data, execute deep crawls, and budget credit costs.

Authentication & Multi-Tenant Usage

datamint tools require a valid API key (mint_live_... or mint_test_...). You can provide your key in two ways:

Per-Tool Parameter (Hosted Multi-Tenant)api_key: "mint_live_..."

When calling the shared hosted endpoint https://mcp.datamint.xyz/sse, your agent or LLM can pass its API key in the tool arguments. All tools accept an optional api_key parameter.

Environment Variable (Local Stdio & Self-Hosted)DATAMINT_API_KEY="mint_live_..."

When running locally via npx @datamint/mcp, define DATAMINT_API_KEY in your client configuration or shell environment. Tools will use this key automatically.

Need an API key? You can generate one instantly with 1,000 free starter credits on your dashboard.

Exposed Tools

The server registers 5 native tools conforming strictly to the Model Context Protocol specification:

datamint_mintCost: 1–17 credits

Single-page synchronous retrieval. Returns cleaned, noise-free Markdown or structured JSON from any URL.

Parameters

url(string)requiredTarget web URL (http/https)

js(boolean)optionalHeadless Chrome render for React/Vue SPAs (5 credits)

stealth(boolean)optionalResidential proxy routing + CAPTCHA bypass (15 credits)

schema(string | object)optionalRegistered schema name or custom JSON Schema (+2 credits)

include_screenshot(boolean)optionalAttach base64 screenshot image to MCP response

api_key(string)optionalOptional API key override for multi-tenant hosted usage

datamint_extractCost: 3–17 credits

Zero-code extraction into validated JSON conforming strictly to a canonical schema or custom JSON Schema.

Parameters

url(string)requiredTarget web URL

schema(string | object)requiredSchema name (ecommerce_item, article, job_posting, etc.) or JSON Schema object

js(boolean)optionalHeadless Chrome render

stealth(boolean)optionalResidential proxy routing

api_key(string)optionalOptional API key override

datamint_crawlCost: 1–15 credits / page

Asynchronous deep crawl traversing domains and subdomains with path regex filters and depth limits.

Parameters

url(string)requiredSeed root URL to crawl

max_depth(number)optionalLink hop depth (0–10, default: 2)

max_pages(number)optionalPage budget cap (1–1,000, default: 25)

path_include(string)optionalRegex pattern for required URL paths

path_exclude(string)optionalRegex pattern for excluded URL paths

stealth(boolean)optionalRoute all requests via residential proxy pool

webhook_url(string)optionalHTTPS callback URL notified upon crawl completion

datamint_crawl_statusCost: 0 credits (free)

Poll crawl status, inspect progress counters (pages done, failed, credits used), and fetch extracted pages.

Parameters

crawl_id(string)requiredUUID returned by datamint_crawl

limit(number)optionalNumber of page records to fetch (1–100)

offset(number)optionalPagination offset

datamint_estimate_costCost: 0 credits (free)

Pre-flight calculation of exact credits and USD cost ($1.00 USD = 1,000 credits) before running workloads.

Parameters

workload_type(mint | extract | crawl)requiredWorkload operation type

count(number)optionalNumber of target pages (default: 1)

js(boolean)optionalHeadless browser render

stealth(boolean)optionalResidential proxy routing

schema(boolean)optionalSchema extraction pass

Canonical Schemas & Resources

The MCP server exposes registered resources that agents can inspect for pre-flight context or read directly:

datamint://schemasResource catalog

Catalog of pre-mapped canonical extraction schemas and field documentation.

datamint://schemas/{name}JSON Schema

Returns the strict JSON Schema definition for a specific schema model.

datamint://system/statusFleet telemetry

Live crawler fleet metrics: active headless browser workers, queue latency, and proxy pool status.

datamint://crawls/{id}Crawl records

Direct URI for agents to stream extracted pages from a finished or in-progress crawl job.

Built-in Canonical Models

ecommerce_item

Product title, price, currency, availability, rating, reviews, SKU, images

article

Headline, author, published date, canonical URL, full body Markdown, summary

job_posting

Role title, company, location, salary range, employment type, requirements

company_profile

Company name, domain, industry, headquarters, employee count, description

real_estate_listing

Address, price, bedrooms, bathrooms, square footage, property type

faq

Extracted question-and-answer pairs across help centers and landing pages

Guided Agent Prompts

In addition to tools, the MCP server provides prompt templates that guide LLMs through complex web data workflows:

research_target

Analyzes a target URL with automatic tier escalation. Starts with fast static retrieval, evaluates challenge signals, and escalates to dynamic rendering or residential stealth proxies as needed.

extract_structured_data

Guides the agent to inspect the schema catalog, select or validate an extraction schema, and invoke datamint_extract with confidence score evaluation.

crawl_knowledge_base

Plans and executes a multi-page documentation or product catalog crawl, running pre-flight cost estimation with datamint_estimate_cost before launching the job.

Engine Architecture: Transparent Escalation

When your agent calls datamint_mint, it does not need to worry about Cloudflare Turnstile, DataDome, Kasada, or headless browser memory leaks. The datamint ingestion pipeline automatically executes transparent escalation:

Tier 1: Fast JA4/TLS Emulation1 credit · ~800ms

Direct HTTP/2 request with Chrome 133 TLS handshake fingerprinting. Bypasses passive bot detection with zero browser overhead.

Tier 2: Headless Browser Engine5 credits · ~2.5s

Full DOM execution, JavaScript hydration, canvas rendering, and CSS selector wait states for SPAs.

Tier 3: Stealth Armor & Residential Routing15 credits · ~4.5s

Rotates through residential IP pools, mutates canvas/WebGL fingerprints, and executes autonomous CAPTCHA click-solving.

Official Container (Docker / GHCR)

For self-hosted deployments, Kubernetes clusters, or isolated agent sandboxes, the official container image is published to GitHub Container Registry at ghcr.io/dataminthq/datamint-mcp:latest.

# Pull the official container image
docker pull ghcr.io/dataminthq/datamint-mcp:latest

# Run interactively via stdio
docker run -i --rm -e DATAMINT_API_KEY="mint_live_xxx" ghcr.io/dataminthq/datamint-mcp:latest

# Or run as a background SSE service (port 3001)
docker run -d --name datamint-mcp -p 3001:3001 \
  -e DATAMINT_API_KEY="mint_live_xxx" \
  ghcr.io/dataminthq/datamint-mcp:latest \
  node bin/datamint-mcp.js --transport sse --port 3001

Verifying Endpoint Health

You can test connectivity to our MCP server at any time using standard HTTP tools:

# Check server health
curl -s https://mcp.datamint.xyz/health

# Output:
# {"status":"ok","service":"datamint-mcp","version":"0.1.0"}

# Test SSE stream initiation
curl -N -H "Accept: text/event-stream" https://mcp.datamint.xyz/sse

Need volume pricing or a custom parser?

Start free Talk to sales