Blopus is a web search API for AI agents and LLMs — one REST call, or a one-line web search MCP server for Claude, Cursor and any MCP client. Every result comes from our own crawler and index, never resold, so your agent can search the web, read the page, and act on what's true right now.
// Add Blopus to Claude, Cursor, or any MCP client { "mcpServers": { "blopus": { "url": "https://mcp.blopus.ai", "headers": { "Authorization": "Bearer blp_live_..." } } } }
curl https://api.blopus.ai/v1/search \ -H "Authorization: Bearer blp_live_..." \ -H "Content-Type: application/json" \ -d '{"query":"site reliability trends 2026","count":5,"freshness":"pm"}'
import requests r = requests.post("https://api.blopus.ai/v1/search", headers={"Authorization": "Bearer blp_live_..."}, json={"query": "latest on vector databases", "count": 5}) for x in r.json()["results"]: print(x["title"], x["url"])
const r = await fetch("https://api.blopus.ai/v1/search", { method: "POST", headers: { Authorization: "Bearer blp_live_...", "Content-Type": "application/json" }, body: JSON.stringify({ query: "who won the match last night", count: 5 }), }); const { results } = await r.json(); console.log(results[0].snippet);
Raw HTML is noise. Blopus returns ranked results with clean, de-duplicated, citation-ready snippets — from an index we own and keep fresh, so your context window works for you, not against you.
Our own crawler and index. New pages are searchable within minutes, and we never resell Google or Bing. No third-party terms, quotas, or price changes sitting between your agent and the web.
Hybrid semantic + keyword ranking finds results by meaning and exact terms, returns clean model-ready snippets, and connects over native MCP. ~130 ms typical.
~130 ms typical server-side latency; end-to-end varies with your network, region, and connection.One price, a hard cap, zero overage. Hit your monthly searches and calls pause until reset — a runaway agent can never run up a surprise bill. Top up any time if you need more.
One key, one endpoint. Tap your tool for a copy-paste snippet — MCP, function calling, an SDK, or plain REST — plus a link to its official docs.
Plus any MCP client or tool-calling agent. If your model or framework can call a tool — natively over MCP, or as a function in any agent loop — Blopus drops right in. Cline, Windsurf, Continue, LangChain, LlamaIndex, CrewAI, AG2, n8n, and hundreds more. One key, one endpoint, everywhere — no SDK, no lock-in.
Blopus runs its own crawler and search index end to end. We don't resell Google or Bing, and we don't scrape other search engines — so you get an independent source of truth that stays up when someone else's terms, quotas, or prices change. New pages become searchable within minutes.
Connect once and every tool-using model can search, fetch, and cite the live web. No custom tool code, no browser farm, no maintenance. Point Claude, Cursor, or your own runtime at Blopus and your agents get more reach — instantly. Bundled in every plan.
https://mcp.blopus.ai — nothing to deploy, patch, or maintain.search and fetch — with freshness, domain filters, date ranges, relevance scores, pagination, and batch fetch built in (see below).{
"mcpServers": {
"blopus": {
"url": "https://mcp.blopus.ai",
"headers": { "Authorization": "Bearer blp_live_..." }
}
}
}# any MCP-aware agent framework tools = mcp.connect("https://mcp.blopus.ai", key=KEY) agent = Agent(model="your-llm", tools=tools) agent.run("Research 3 vendors and cite sources.") # the model now calls search / fetch on its own
search + fetch — with the controls production agents actually need.The surface stays tiny so any model can drive it, but every knob you'd reach for is a parameter away. All live in the API and over MCP today.
Syndicated copies and near-identical template pages are collapsed server-side — one story per slot, with a duplicate_count showing what was merged. No spam eating your results or your tokens.
Set include_content and search returns the cleaned page text inline — enough to read and reason on without a second fetch. One call, fewer round-trips, less latency.
Bias results toward the newest pages, or scope to a window with freshness — pd, pw, pm, py. Built for time-sensitive and news queries.
Narrow to the sources you trust or shut out the ones you don't with include_domains and exclude_domains.
Pin results to an exact span with start_date and end_date — perfect for research over a known period.
Every result carries a score from 0–1, so your agent can threshold, sort, or drop weak matches before spending context.
Page deeper with offset, and let more_results in the response tell the agent when to keep going.
Ask for extended snippets inline and skip a whole fetch round-trip when a richer preview is all you need.
Pass many URLs to fetch in a single call and get clean page text back for all of them — billed per URL.
Constrain results by language, and read your remaining quota echoed on every response — no separate usage call.
One flat monthly price for a fixed number of searches. Hit your cap and calls pause until it resets or you upgrade — we never send an overage bill. Need a little more? Buy a one-time top-up. That's the whole point.
Who is this license for?
This is the licensee, so it goes on the license and on every invoice. Your billing address and, if you have one, a VAT or tax ID are collected at checkout.
Want to try first? 400 free searches — no card, no expiry.
Effective ≈ $1.65–2.33 per 1,000 searches · flat rate, no metered overage. Hit the cap and buy a one-time top-up — 1,000 searches for $5. How searches and credits are counted.
Prices in USD, excluding sales tax where applicable. Personal plans are licensed to you and cover your own work — your login and API keys are yours alone and cannot be shared, transferred or sold. Building something other people use? See business plans.
Two questions to start, and it goes straight to a Blopus representative who will reply within one business day.
Most search APIs hand your model raw links and a metered bill. Blopus ships the whole agent toolkit — hybrid search, full-page fetch, and a native MCP server — over an index we own, with pricing you can actually cap.
Capability table, monthly-cost table for every tier, and a vendor-by-vendor breakdown — Brave, Serper, Tavily, Exa, SerpAPI and Perplexity — with every price read from that vendor’s own page and date-stamped: blopus.ai/compare.
Metered APIs punish you the moment an agent loops. Blopus is flat-rate and hard-capped — you pay your plan, full stop. That's the entire pitch.
Pay your monthly plan and nothing else. No per-call metering, no usage invoice, no end-of-month math.
Reach your monthly searches and calls pause until reset or upgrade. We never charge a cent over your plan — a runaway agent can't run up a tab. Buy a one-time top-up if you need more.
Live usage dashboard with alerts at 50 / 80 / 100%, so you always know where you stand.
Everything you'd want to know before wiring Blopus into your agents. Still stuck? Email us.
blp_live_... — drop one into the REST API, a Python/TS SDK, or the hosted MCP endpoint and you're reading the live web in one line.https://mcp.blopus.ai — point Claude, Cursor, ChatGPT, or any MCP / OpenAI-tool client at it with your key and the search and fetch tools appear automatically. Nothing to deploy.Grab a key, paste one snippet, and your model is reading the live web — no SDK, no infrastructure, nothing to deploy.