AI Deep Research — Multi-Source Research API for AI Agents, RAG & Grounding (Keyless)

Give it a topic, get everything a deep-research agent needs to reason and cite. AI Deep Research is a fast, keyless deep-research API built for AI agents, RAG pipelines and grounding. Send one topic or a batch of fifty and, for each, receive a clean, structured, cited research bundle: the top-ranked web sources, the full readable content of every source as Markdown, and recent news — all in one call. No API key, no browser, no rate-limit headaches, no per-token synthesis fee.

Crucially, this Actor does not synthesize. There is no LLM inside it and no opinion baked in — it returns clean, cited source material so your agent's own model can read, reason and write the answer. Think of it as the retrieval-and-reading layer of a deep-research stack: it does the searching, fetching, deduping and Markdown conversion; your LLM does the thinking. Large language models have a knowledge cutoff and hallucinate about anything recent or niche — this Actor fixes that by handing your agent fresh, real sources with full text and citations instead of guesses.

🏆 Why this deep-research API?

3 signals per topic (web search + full-page Markdown + recent news) · up to 20 sources each · batch up to 50 topics per run · ~6,000 chars of clean Markdown per source · citations out of the box · resilient dual engine (DuckDuckGo → Bing fallback) · keyless with Apify proxy · MCP / pay-per-event so agents can call it autonomously. The unofficial Tavily / Exa / Perplexity API alternative for grounded retrieval — without the API key or the synthesis lock-in.

---

✨ What this Actor does / Key features

• 🔑 Keyless & cheap — no Tavily, Exa, Perplexity, Brave or SerpAPI key required. Runs on plain HTTP through Apify's proxy.
• 🧩 Multi-source by design — every topic combines a web search + the full page content of each result + recent news, the three ingredients a research agent needs, in one record.
• 📖 Deep, not shallow — turn on Include source content (on by default) and each source carries the page's main text as clean Markdown (capped ~6,000 chars). Your LLM reads real sources, not one-line snippets.
• 🔖 Cited out of the box — every bundle ships a citations list ({n, title, url}) so the agent can footnote its answer.
• 🛡️ Resilient dual engine — DuckDuckGo HTML is primary; if it returns nothing, Bing takes over automatically, so you get sources, not empty responses. The engine field records which one served.
• 📰 Fresh news — recent items from Google News RSS (last 30 days): title, URL, publish date, source.
• 🤖 Agent-native (MCP) — pay-per-event pricing means an AI agent can research a topic, pay per bundle, and reason over the output with no account setup.
• 🌍 Region / language control — us-en, uk-en, de-de, fr-fr, es-es and more.
• 📚 Batch — fire up to 50 topics in one run with configurable concurrency.
• 💾 Export-ready — JSON, CSV, Excel and a full REST API. One row per topic.

🚀 Quick start (3 steps)

1. Configure — type one or more topics into Research topics (or a single topic into Single topic). Optionally set Max sources per topic and a Region / language.
2. Run — click Start. The Actor searches the live web, fetches and converts each source to Markdown, and attaches recent news.
3. Get your data — open the Output tab: one cited research bundle per topic. Export to JSON, CSV or Excel, read it live via the Apify API, or call the Actor as an MCP tool from your agent.

📥 Input

Provide at least one topic — either in the Research topics list or the Single topic field. Everything else is optional.

Example — batch research for a report agent

{
  "topics": [
    "model context protocol adoption",
    "vector database market 2026",
    "AI agent orchestration frameworks"
  ],
  "maxSources": 8,
  "includeContent": true,
  "includeNews": true,
  "maxNews": 8,
  "region": "us-en",
  "concurrency": 2
}

Example — single deep dive with maximum source content (RAG grounding)

{
  "topic": "post-quantum cryptography standards NIST",
  "maxSources": 12,
  "includeContent": true,
  "includeNews": false,
  "region": "us-en"
}

Example — fast news-and-links scan (no full page fetch)

{
  "topics": ["OpenAI funding round", "Anthropic enterprise deals"],
  "maxSources": 6,
  "includeContent": false,
  "includeNews": true,
  "maxNews": 15,
  "region": "us-en",
  "concurrency": 4
}

Field	Type	Description
topics	array	One or more topics / questions. Each produces one cited research bundle. Batch up to 50.
topic	string	A single topic / question. Use this or the topics list above.
maxSources	integer	Top-ranked web sources to gather per topic (1–20). Default 6.
includeContent	boolean	Fetch each source page and attach its main content as clean Markdown (~6,000 chars). This is what makes it "deep". Default true.
includeNews	boolean	Attach recent news from Google News RSS (last 30 days). Default true.
maxNews	integer	Recent news items to attach per topic (0–30). Set 0 to skip news. Default 8.
region	string	Search region / language code, e.g. us-en, uk-en, de-de, fr-fr. Default us-en.
concurrency	integer	Topics researched in parallel (1–8). Kept low by default because each topic runs many fetches. Default 2.
useApifyProxy	boolean	Route requests through Apify datacenter proxy to avoid per-IP rate limits. Default true.
proxyGroups	array	Override proxy group, e.g. ["RESIDENTIAL"]. Leave empty for datacenter (AUTO).

Tip: Keep includeContent on for RAG and report-writing (your LLM reads full sources); turn it off for a lightweight links-and-news scan that runs faster and cheaper. Each source adds one page fetch, so a lower concurrency keeps runs stable when maxSources is high.

📤 Output

One row per topic — a self-contained, cited research bundle. Here is a trimmed sample record (long arrays and content truncated with …):

{
  "topic": "model context protocol adoption",
  "sources": [
    {
      "rank": 1,
      "title": "Model Context Protocol — An open standard for AI tools",
      "url": "https://modelcontextprotocol.io",
      "snippet": "MCP standardizes how applications provide context to LLMs…",
      "source": "modelcontextprotocol.io",
      "content": "# Model Context Protocol\n\nMCP is an open protocol that standardizes how applications provide context to large language models…",
      "contentWordCount": 812
    },
    { "rank": 2, "title": "…", "url": "https://…", "snippet": "…", "source": "…", "content": "…", "contentWordCount": 640 }
  ],
  "news": [
    {
      "title": "Anthropic's MCP gains traction across IDEs",
      "url": "https://news.example.com/mcp",
      "publishedAt": "Wed, 02 Jul 2026 09:00:00 GMT",
      "source": "Example News"
    }
  ],
  "citations": [
    { "n": 1, "title": "Model Context Protocol — An open standard for AI tools", "url": "https://modelcontextprotocol.io" },
    { "n": 2, "title": "…", "url": "https://…" }
  ],
  "sourceCount": 6,
  "newsCount": 8,
  "engine": "duckduckgo",
  "generatedAt": "2026-07-04T10:00:00.000Z"
}

💡 Use cases

• Deep research agents — the retrieval + reading layer for an agent that writes long, cited reports. It gathers and reads sources; your LLM synthesizes.
• RAG & grounding — drop full-text, cited sources straight into a prompt or vector store instead of stale training data.
• Research automation — batch dozens of questions and collect ranked, read sources plus fresh news in one run.
• Market & competitive intel — one topic → sources + recent news, ready to summarize.
• Answer engines & chatbots — power a bot that answers with links and citations, not hallucinations.
• Fact-checking & briefings — pull the current top sources and last-30-days news on any subject on demand.

👥 Who uses it

AI engineers & agent builders · RAG / LLMOps teams · researchers & analysts automating literature and market scans · product teams building answer engines and research assistants · founders and consultants who need fast, cited briefings · anyone replacing a paid search/grounding API with a keyless, pay-as-you-go alternative.

🤖 Use it as an MCP tool

This Actor uses Apify's pay-per-event model, so AI agents can discover and call it autonomously through the Apify MCP server — one charge per topic bundle. Wire it into Claude, Cursor, or any MCP client and let the agent research the live web, read the sources, and cite them on demand — no account setup, no key handling.

💰 Pricing

This Actor runs on a simple pay-per-result (pay-per-event) model — one charge per topic bundle returned. Topics that yield no sources still push a row but are not charged. There is no monthly fee and no idle cost. Try it on the free tier first, then scale up; see the Pricing tab on this page for the current rate.

❓ Frequently Asked Questions

Is this a Tavily / Exa / Perplexity API alternative?

Yes — it covers the deep-research / grounding job those tools are used for (multi-source retrieval, full-page reading and citations) without their API keys. The difference: it returns raw, cited source material and leaves synthesis to your own LLM, so you keep full control and pay no per-token synthesis fee.

Does it use an LLM or write the answer for me?

No. There is no LLM inside. It searches, reads and cites sources; your research agent's model does the reasoning and writing. That keeps it keyless, deterministic and cheap.

Can I use it without an API key or login?

Yes. It's fully keyless — no Tavily/Exa/SerpAPI key, no third-party signup, only an Apify account. It works out of the box with Apify's datacenter proxy.

Where do the sources come from?

A live web search — DuckDuckGo HTML first, with an automatic Bing fallback for resilience — plus recent news from Google News RSS (last 30 days). The engine field tells you which search served each bundle.

Can it return the full page content, not just snippets?

Yes — that's the point. Include source content (on by default) attaches each source page's main content as clean Markdown (capped ~6,000 chars), ideal for RAG and deep reading.

How do I export the research to CSV or JSON?

Run the Actor, then export the dataset as JSON, CSV, Excel or JSONL from the Apify console or via the REST API. One topic = one row.

How much data can I get per run?

Up to 50 topics per run, each with up to 20 sources and 30 news items. With includeContent on, every source carries ~6,000 chars of Markdown — so a single run can return a large, fully-cited research corpus.

Can I research in other languages/regions?

Yes. Set the Region / language field, e.g. uk-en, de-de, fr-fr, es-es.

Is it legal to use?

The Actor retrieves only publicly available web pages and news. You are responsible for using the returned content in compliance with each source's terms and applicable law (including copyright and data-protection rules).

Why is the default concurrency low?

Each topic runs many fetches (a search, one fetch per source, and news), so a lower concurrency keeps runs stable and friendly to sources. Raise it if you have many light topics or turn includeContent off.

🔗 More AI & research-intel Actors by logiover

Building an AI research or grounding stack? Pair AI Deep Research with the rest of the suite:

Actor	What it does
AI Web Search	Keyless live web search results for agents
AI Web Extract	Turn any URL into clean, LLM-ready content
AI Citation Source Finder	Find citable sources for any claim
Company Deep Research Scraper	Multi-source company research dossiers
News Intelligence Scraper	Aggregated, structured news feeds
Google News Scraper	Scrape Google News results at scale
SERP Keyword Research	Keyword and SERP data for research
arXiv Paper Scraper	Fetch academic papers and metadata from arXiv
Semantic Scholar Research Scraper	Academic papers, citations and authors
Hugging Face Hub Intelligence Scraper	Models, datasets and Spaces intelligence
GitHub Repository Scraper	Repo metadata, stars and activity
npm Package Intelligence Scraper	Package stats, deps and release data
CVE Security Advisory Monitor	Track new CVEs and advisories
B2B AI Visibility Tracker	Measure brand visibility in AI answers

👉 Browse all logiover scrapers on Apify Store — 180+ actors across real estate, jobs, crypto, social media & B2B data.

⏰ Scheduling & integration

Schedule this Actor on Apify to refresh a research corpus daily or weekly and keep your grounding data current. Export results to JSON, CSV or Excel, sync to Google Sheets, or push straight into your vector store, database, BI tools and webhooks through the Apify API. Connect it to Make, n8n or Zapier to build automated research pipelines — or call it as an MCP tool directly from an AI agent.

⭐ Support & feedback

Need an extra field or a different engine? Open an issue on the Issues tab — response is usually fast. If this Actor saves you time, a ★★★★★ review on the Store page genuinely helps and is hugely appreciated. 🙏

⚖️ Legal

This Actor retrieves only publicly available web pages and news items and is intended for legitimate research, grounding and analytics use. It does not synthesize or alter source content. You are responsible for using the returned material in compliance with each source's terms of service, copyright and any applicable data-protection laws.

---

📝 Changelog

2026-07-06

• ✨ README overhaul: rising-star formatting, badge row, named example scenarios, full field reference, expanded keyword-gap FAQ and AI-research cross-promo grid.

2026-07-04

• Initial release: keyless multi-source deep-research API for AI agents and RAG.
• Per topic: runs a web search (DuckDuckGo HTML primary, Bing automatic fallback) and gathers the top maxSources ranked sources — rank, title, URL, snippet, source domain.
• Optional includeContent (on by default) — fetches each source page and attaches its main content as clean Markdown (capped ~6000 chars) so the agent reads full sources, not just snippets. Fetched in parallel per topic.
• Optional includeNews (on by default) — attaches recent news from Google News RSS (last 30 days): title, URL, publish date, source.
• Assembles one cited record per topic with a citation list ({n, title, url}), sourceCount and newsCount.
• Does NOT synthesize (no LLM) — returns clean, structured, cited source material the agent's own LLM reasons over.
• Region/language control; batch topics with configurable (low) concurrency since each topic runs many fetches.
• Keyless; Apify datacenter proxy by default (RESIDENTIAL opt-in).
• Pay-per-event pricing: one charge per topic bundle. Topics that yield no sources still push a row (no charge). Agent-first, works as an MCP tool.