Datasets for training. Search, scrape, proxy, and browser infrastructure for reaching any live data.
One platform, one API key — powered by our own network across 195 countries.
graph LR
N(("Novada<br/>one API key")):::hub
N --> DATA["🧠 Datasets<br/>AI · Embodied AI"]:::data
N --> REACH["🌐 Reach any data<br/>Search · Scrape · Proxy<br/>Browser · Monitor · Verify"]:::reach
N --> SHIP["🚀 Delivered as<br/>Hosted MCP · 15 tools<br/>npm · 23 tools · SDKs"]:::ship
DATA --> OUT(["Your models & agents"]):::out
REACH --> OUT
SHIP --> OUT
classDef hub fill:#011627,color:#ffffff,stroke:#011627;
classDef data fill:#44289F,color:#ffffff,stroke:#44289F;
classDef reach fill:#F5F4FA,color:#011627,stroke:#E7E7E8;
classDef ship fill:#F5F4FA,color:#011627,stroke:#E7E7E8;
classDef out fill:#ffffff,color:#011627,stroke:#44289F;
Trusted by 8,000+ companies worldwide — we own the network, no reseller markup.
 Search & SERP 5-engine web search · SERP analytics · deep research · live fact-checking |
 Scraping & Extraction 129 structured platforms · any URL → JSON/markdown · async pipeline |
 Proxy Network 100M+ IPs · 6 types · 195+ countries · we own the network |
 Browser & Unblock CDP cloud browser · anti-bot bypass · JS rendering · session automation |
 Monitor & Verify Page change detection · claim fact-checking · diff alerts |
 AI Brand Monitor ChatGPT · Perplexity · Claude · Gemini · Grok — mentions & sentiment |
100M+ residential IPs across 195+ countries — city-level targeting, on every continent.
Text taught models to talk. The next generation needs the video, real-world, and web-scale data it takes to act. Novada maintains a large, continuously-refreshed catalog of training-ready datasets — and sources custom ones on request.
|
Curated, ready-to-train datasets for LLMs and multimodal models.
|
Grounded, real-world datasets for perception and world models.
|
| 📝 Text | 🖼️ Images | 🎬 Video | 🤲 Real-world interaction | 📍 Geo / location | 🛒 Commerce |
|---|
Most vendors sell you a proxy. Novada is the data layer that trains the AI — the proxy is included, not the headline.
graph LR
A["🌐 Live Web<br/>Real World"]:::n --> B["⚙️ Novada<br/>collect · unblock<br/>structure · package"]:::hub
B --> C["✨ Clean data<br/>datasets · live streams"]:::n
C --> D["🤖 Your models<br/>& agents"]:::out
classDef n fill:#F5F4FA,color:#011627,stroke:#E7E7E8;
classDef hub fill:#44289F,color:#ffffff,stroke:#44289F;
classDef out fill:#ffffff,color:#011627,stroke:#44289F;
One call in → routed across datasets, proxies, browsers, and scrapers → clean data or reliable access lands in your pipeline, ready to train or act on.
Get your key at novada.com — free tier, no credit card.
Option A — Hosted MCP (zero install, 15 tools):
{
"mcpServers": {
"novada": { "url": "https://mcp.novada.com/mcp?token=your_key" }
}
}# Claude Code
claude mcp add --transport http novada "https://mcp.novada.com/mcp?token=your_key"
# Or keep your key out of the URL with a Bearer header:
claude mcp add --transport http novada https://mcp.novada.com/mcp --header "Authorization: Bearer your_key"Path style https://mcp.novada.com/<your_key>/mcp also works.
Option B — Local npm (all 23 tools, including browser automation):
# Claude Code
claude mcp add novada -e NOVADA_API_KEY=your_key -- npx -y novada-mcp{
"mcpServers": {
"novada": {
"command": "npx",
"args": ["-y", "novada-mcp"],
"env": { "NOVADA_API_KEY": "your_key" }
}
}
}All 23 MCP tools — click to expand
| Tool | What it does |
|---|---|
| Content retrieval | |
novada_search |
5-engine web search (Google/Bing/DuckDuckGo/Yandex) with geo-targeting and optional inline extraction. |
novada_extract |
Any URL (up to 10) → clean content, links, and structured fields. Auto static/render/browser escalation. |
novada_crawl |
BFS/DFS site walk with configurable depth — extract content from each page. |
novada_research |
Multi-step research: 3–10 parallel queries → dedup → extract → synthesized report. |
novada_map |
Discover a site's full URL structure without reading content. |
novada_site_copy |
Copy an entire docs site to disk as one markdown file per page. |
novada_search_feedback |
Record search-result quality to bias future ranking. |
| Scraping & verification | |
novada_scrape |
16 structured platforms, ~88 operations (Amazon, TikTok, LinkedIn…) → structured records. |
novada_ai_monitor |
How ChatGPT, Perplexity, Grok, Claude, Gemini mention a brand — mentions, sentiment, positioning. |
novada_monitor |
Page change detection between calls. |
novada_verify |
Fact-check a claim via 3 parallel search angles. Returns verdict + confidence. |
| Proxy | |
novada_proxy |
Proxy credentials for your own HTTP clients — 6 types: residential, ISP, datacenter, mobile, static, dedicated. |
| Browser (local npm only) | |
novada_browser |
Session-persistent CDP cloud browser — navigate, click, type, screenshot, evaluate. |
novada_browser_flow |
Multi-step browser automation via an action-sequence API. |
| Account & billing | |
novada_account |
Unified dashboard — wallet balance, plan quotas, usage. |
novada_proxy_account_create |
Create a proxy sub-account (write, confirm-gated). |
novada_proxy_account_list |
List proxy sub-accounts for a product. |
novada_ip_whitelist |
Manage the proxy IP whitelist (add/list/remove). |
novada_capture_apikey |
Get or reset the Capture API key. |
novada_static_ip_mgmt |
Manage static ISP IPs — open, renew, export, list. |
| Health & discovery | |
novada_discover |
List all available Novada tools with descriptions and categories. |
novada_setup |
First-run concierge — validates your API key and orients you. |
novada_session_stats |
Per-session usage telemetry — call counts and uptime. |
Full reference at developer.novada.com.
- We own the network. 100M+ residential IPs across 195 countries — no reseller markup, no third-party SLAs.
- Datasets, not just access. A large, continuously-refreshed catalog for AI and embodied-AI training — plus custom sourcing on request.
- One key, everything unlocked. Datasets, search, scrape, proxy, monitor, browser, and AI brand tracking — no per-product billing surprises.
- Built for agents and developers alike. Native MCP server for AI agents, Python and Go SDKs for direct integration, REST API for everything else.
| AI Clients | Agent Frameworks | Automation |
|---|---|---|
| Claude Desktop · Claude Code | OpenAI Agents SDK | n8n |
| Cursor · Windsurf · Cline | LangChain | Zapier (coming soon) |
| VS Code | LlamaIndex · CrewAI | |
| Codex CLI |
Full integration guides at developer.novada.com.
| If you need… | Use |
|---|---|
| Everything — zero install, hosted | https://mcp.novada.com/mcp?token=your_key |
| Everything — local, all 23 tools + browser | npx novada-mcp |
| Search, scrape, crawl, research only | npx novada-search |
| Proxy credentials only | npx novada-proxy-mcp |
Not sure? Start with the hosted endpoint — same key, zero setup.
Novada is a hosted platform — start at novada.com with one API key. Full docs at developer.novada.com.
mcp.novada.com — remote Streamable-HTTP endpoint, 15 tools. Paste one URL. Always up to date — tools update server-side, no client redeploy needed.
https://mcp.novada.com/mcp?token=your_key
novada-mcp — all 23 tools, including browser automation.
npx novada-mcp
novada-search-mcp — search, extract, crawl, research · npm: novada-search
novada-proxy — 6 proxy types for AI agents · npm: novada-proxy-mcp
- novada-python — official Python client for the Novada API.
- novada-go — official Go client for the Novada API.
- novada-proxy-extension — route browser traffic through Novada's proxy network (Chrome, Manifest V3).
- novada-scraper-skill — agent skill: turn any website into structured data.
- novada-webunblocker-skill — agent skill: reach sites that fight back.
Coming soon: deeper LangChain · CrewAI · n8n · Zapier integrations.
Building with Novada? We're happy to integrate with any agent framework, MCP client, or automation platform. Open an issue or reach us at novada.com.
The data layer for AI — for the people building with it, and the agents they build.
Get an API key · Hosted MCP · Star the flagship · Read the docs
