Native MCP tool — works inside Claude, Cursor & more

Autonomous research
for B2B sales teams

Not a search API. Not a $500/mo platform. An autonomous AI agent that searches 20+ sources, scrapes LinkedIn & JS-heavy pages, and returns structured JSON — all for 1 credit per query.

See it in action

200 free credits

No credit card

Setup in 2 min

Claude Desktop

Research Stripe — find CEO, employee count, latest funding round, and main competitors. Return as JSON.

Calling research tool...

{
  "ceo": "Patrick Collison",
  "employees": 8000,
  "latest_funding": "$6.5B Series I",
  "valuation": "$50B",
  "competitors": [
    "Adyen", "Square", "Braintree"
  ]
}

12.3s 23 sources 1 credit

Sources per query

Avg response time

AI models supported

MCP tools available

How It Works

From question to structured data
in under 30 seconds

Not a search API. Not a scraper. A fully autonomous agent that plans, searches, scrapes, cross-references, and synthesizes — all in one query.

You Ask

Ask in natural language or via API. "Find the CEO, funding, and ICP of Stripe."

AI Searches

Searches Google, scrapes websites, LinkedIn, Crunchbase — even JS-rendered pages.

AI Reasons

Cross-references 20+ sources, validates data, resolves conflicts, and structures the output.

You Get JSON

Structured, validated data in your exact schema. Ready for your CRM, spreadsheet, or pipeline.

Why not just use...

We sit in the whitespace between raw APIs and expensive platforms. Here's why that matters.

vs Clay

$185/mo for static data

Clay pulls from pre-indexed databases with $185/mo minimum. We research the live web — deeper, fresher, and pay only for what you use.

vs Tavily / Exa

Search results aren't research

Search APIs return raw links. We run autonomous multi-step research: plan, search, scrape, cross-reference, and synthesize into structured JSON.

vs Relevance AI

Generic agents, manual setup

Agent builders need hours of configuration. We're purpose-built for B2B research — pre-built templates, zero setup, native MCP integration.

vs Manual

15 min per lead doesn't scale

An SDR spends 15 minutes per lead. We return structured, validated data in under 30 seconds — at €0.024 per query.

Reusable Agents

Configure once, research 1,000 leads

Create autonomous research agents with custom prompts, templates, and output schemas. Run via API, MCP in Claude, or dashboard.

ICP Qualification Agent

google/gemini-2.5-flash

Active

Prompt Template

Research {{company_name}} at {{website}}. Find:
- Company size and headcount
- Industry and target market
- Technology stack
- Recent funding or news
- Decision makers (C-level)

Input Variables

company_name website

Output Schema

{ "employees": number, "industry": string,
  "funding": string, "ceo": string, ... }

Pre-built templates to get started fast

Start from a template or build your own. Each agent has a prompt, input variables, and optional output schema for structured JSON.

ICP Qualification

Check if a company matches your ideal customer profile based on size, industry, tech stack.

Company Enrichment

Full company profile: headcount, revenue, funding, leadership, tech stack, competitors.

Lead Research

Research a person: role, LinkedIn, recent posts, company context, talking points.

Competitive Analysis

Compare a company against competitors: positioning, pricing, features, market share.

Bulk Runs

Run any agent on hundreds of leads at once. Upload a list, get structured results for all.

Live Demo

Try it yourself

See how Research Agent works with different integration methods.

# Research any company with a single API call
curl -X POST https://research-agent.net/research \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d prompt="Find the CEO, employee count, and funding for Stripe" \
  -d model="google/gemini-2.5-flash" \
  -d schema='{"ceo":"string","employees":"number","funding":"string"}'

# Response:
# {
#   "output": { "ceo": "Patrick Collison", "employees": 8000, "funding": "$6.5B" },
#   "model_used": "google/gemini-2.5-flash",
#   "execution_time_seconds": 12.3
# }

# pip install requests
import requests

response = requests.post(
    "https://research-agent.net/research",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    data={
        "prompt": "Find the CEO, funding, and employee count for Stripe",
        "model": "google/gemini-2.5-flash",
        "schema": '{"ceo":"string","employees":"number","funding":"string"}',
    }
)

data = response.json()
print(data["output"])
# {"ceo": "Patrick Collison", "employees": 8000, "funding": "$6.5B"}

// Add to Claude Desktop or Cursor MCP settings
{
  "mcpServers": {
    "research-agent": {
      "type": "streamable-http",
      "url": "https://research-agent.net/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

// For Claude.ai web (no headers needed):
// URL: https://research-agent.net/mcp/YOUR_API_KEY

Built for the hardest research problems

Not a simple scraper. A multi-step AI research engine with real browser access.

Deep Multi-Step Research

AI reasons across 20+ sources per query. Follows leads, cross-references data, resolves conflicts. Real research, not a single API call.

LinkedIn & JS Pages

Headless browser infrastructure reaches LinkedIn profiles, Crunchbase, SPAs, and login-gated sites that simple HTTP can't.

Structured JSON Output

Define your own JSON schema. Get exactly the fields you need, in the format your pipeline expects. No parsing needed.

MCP Native

18 tools available via MCP. Use inside Claude Desktop, Claude Code, Cursor, or any MCP-compatible client. Zero setup friction.

100+ AI Models

Pick any model on OpenRouter: GPT-4o, Claude, Gemini, Llama, Mistral. Switch models per query for cost/quality tradeoff.

Bulk Operations

Run agents on hundreds of leads concurrently. Upload a list, track progress, download structured results when done.

Clay depth at Tavily prices

1 credit = 1 autonomous, multi-step research query across 20+ sources. No subscription lock-in.

Clay starts at $185/mo for static enrichments. Tavily gives raw search results at $0.008/call. We give you autonomous deep research from €0.024/query.

Starter

€49

1,000 credits

€0.049 per query

Autonomous multi-source research
LinkedIn + JS page scraping
Structured JSON output
REST API + native MCP

Growth

€149

5,000 credits

€0.030 per query

Everything in Starter
Bulk runs (100s of leads)
Custom agent templates
ICP scoring & buying signals

Scale

€299

12,500 credits

€0.024 per query

Everything in Growth
BYOK — bring your own AI key
Unlimited queries with BYOK
Higher concurrency

All plans include 200 free credits on signup. No credit card required to start.

Frequently Asked Questions

Everything you need to know about Research Agent.

What is a research agent?

A research agent is a reusable AI template that you configure once with a prompt, input variables (like company_name, website), and an optional output schema. You then run it on any number of leads — one at a time or in bulk. The agent searches the web, scrapes pages, and returns structured JSON.

How is this different from just asking ChatGPT or Claude?

Three key differences: 1) We access hard-to-reach pages (LinkedIn, JS-rendered sites, Crunchbase) using headless browsers — chat tools can't. 2) We search and cross-reference 20+ live sources per query, not just the LLM's training data. 3) You get structured JSON output in your exact schema, ready for your CRM or pipeline. Plus, agents are reusable — define once, run on thousands of leads.

What does "1 credit = 1 query" mean?

Each research query or agent run costs exactly 1 credit, regardless of how many sources the AI searches or how long it takes. A query that searches 20+ websites and returns a detailed JSON response costs the same 1 credit. That's €0.024–0.049 per query — compared to Clay at $0.03–0.08 per enrichment plus a $185/mo minimum, or Tavily at $0.008/call for raw search results without synthesis. You get 200 free credits on signup. No hidden costs for web search, scraping, or AI model usage. Scale plan users can bring their own OpenRouter API key for unlimited queries.

Can I run agents on hundreds of leads at once?

Yes! Use the bulk run feature. Submit a list of inputs (e.g., 500 companies with names and websites) and the agent runs on all of them concurrently (5 at a time). You can track progress in real-time and download results when the job completes. Works via API, MCP, or dashboard.

What is MCP and how do I set it up?

MCP (Model Context Protocol) lets AI assistants like Claude and Cursor use external tools natively. Once you add our MCP config (a 6-line JSON snippet), you can just ask Claude: "Research Acme Corp and find their CEO" — and it will call Research Agent automatically, returning structured results directly in the chat. Setup takes under 2 minutes.

Which AI models can I use?

We support 100+ models via OpenRouter, including GPT-4o, Claude Sonnet/Opus, Gemini 2.5 Flash/Pro, Llama, Mistral, and more. The default is google/gemini-2.5-flash which offers the best speed-to-quality ratio. You can switch models per query — use a fast model for bulk enrichment, a powerful model for complex research.

Can I get structured JSON output?

Yes! Pass a JSON schema with your request and the AI will return data matching that exact structure. For example: {"ceo": "string", "employees": "number", "funding": "string"}. This works for both one-off queries and agent runs. Perfect for feeding data directly into your CRM, spreadsheet, or automation pipeline.

Is there an API I can integrate with?

Yes! We offer a full REST API with endpoints for research queries, agent CRUD, bulk runs, billing, and more. Authenticate with your API key via the Authorization: Bearer YOUR_KEY header. Check our API documentation for full details, examples, and parameter schemas.

Legal Notice

Information pursuant to § 5 DDG (German Digital Services Act)

CegTec GmbH

Immenhof 15
22087 Hamburg
Germany

Commercial register: HRB 184866
VAT ID: DE366099274
Registry court: Amtsgericht Hamburg

Represented by Managing Director: Luca Ceglie

Contact

Email: info@cegtec.net
Phone: +49 407 4304626

EU Online Dispute Resolution

The European Commission provides a platform for online dispute resolution (ODR): https://ec.europa.eu/consumers/odr/

We are neither willing nor obliged to participate in dispute resolution proceedings before a consumer arbitration board.

Autonomous research for B2B sales teams

From question to structured data in under 30 seconds