Research & Analysis

Agents that autonomously gather, synthesize, and surface insights from complex data sources.

Research & Analysis Agents

These agents act as tireless research assistants — browsing, reading, cross-referencing, and distilling information into actionable intelligence.

🛠 Active Use Cases

📚 Academic Literature Review Agent

An agent that takes a research question as input, queries academic databases (Semantic Scholar, arXiv), retrieves relevant papers, extracts key findings, and produces a structured literature review with citations. It dramatically reduces the time needed to survey a new field.

import requests
from anthropic import Anthropic

client = Anthropic()

def search_papers(query: str, limit: int = 10) -> list:
    url = "https://api.semanticscholar.org/graph/v1/paper/search"
    params = {"query": query, "limit": limit, "fields": "title,abstract,authors,year"}
    response = requests.get(url, params=params)
    return response.json().get("data", [])

def generate_literature_review(question: str) -> str:
    papers = search_papers(question)
    papers_text = "\n\n".join([
        f"[{p['year']}] {p['title']}\nAuthors: {', '.join([a['name'] for a in p.get('authors', [])])}\nAbstract: {p.get('abstract', 'N/A')[:400]}"
        for p in papers
    ])
    response = client.messages.create(
        model="claude-opus-4-6",
        max_tokens=4096,
        system="You are an academic researcher. Write precise, structured literature reviews with proper citations.",
        messages=[{
            "role": "user",
            "content": f"Research question: {question}\n\nPapers found:\n{papers_text}\n\nWrite a structured literature review."
        }]
    )
    return response.content[0].text

Stack: n8n + Semantic Scholar API + Claude API + Notion

  1. Input: A research topic is submitted via a web form (n8n Webhook trigger).
  2. Search: n8n calls the Semantic Scholar API and retrieves the top 10 papers.
  3. Synthesis: Paper abstracts are sent to Claude to generate a structured review.
  4. Output: The review is saved as a Notion page with bibliography and research domain tags.

🔍 Competitive Intelligence Monitor

An agent that continuously monitors competitor websites, press releases, job postings, and social media to detect strategic signals: new product launches, pricing changes, key hires, and market positioning shifts. It delivers a weekly intelligence brief to the strategy team.

from anthropic import Anthropic
from playwright.sync_api import sync_playwright

client = Anthropic()

def scrape_page(url: str) -> str:
    with sync_playwright() as p:
        browser = p.chromium.launch()
        page = browser.new_page()
        page.goto(url, timeout=15000)
        content = page.inner_text("body")[:3000]
        browser.close()
    return content

def analyze_competitor(company: str, urls: list) -> str:
    all_content = ""
    for url in urls:
        try:
            all_content += f"\n\n[Source: {url}]\n{scrape_page(url)}"
        except Exception as e:
            all_content += f"\n\n[Source: {url}] Error: {e}"

    response = client.messages.create(
        model="claude-sonnet-4-6",
        max_tokens=2048,
        system="You are a strategic analyst. Identify key signals: product changes, pricing, hiring trends, market positioning.",
        messages=[{
            "role": "user",
            "content": f"Analyze this competitor data for {company}:\n{all_content}"
        }]
    )
    return response.content[0].text

Stack: Make + Phantombuster + Claude API + Slack

  1. Schedule: Make scenario runs every Monday morning.
  2. Scrape: Phantombuster agents scrape competitor LinkedIn pages, pricing pages, and blog RSS feeds.
  3. Analyze: Aggregated data is sent to Claude to identify strategic signals.
  4. Deliver: A formatted intelligence brief is posted to a private #competitive-intel Slack channel.

📊 Data Synthesis & Report Generator

An agent that ingests multiple structured and unstructured data sources (CSVs, PDFs, API responses), identifies patterns and anomalies, and generates a coherent executive report with insights and recommendations. It replaces hours of manual analysis.

import pandas as pd
from anthropic import Anthropic

client = Anthropic()

def analyze_dataset(df: pd.DataFrame, report_goal: str) -> str:
    summary = df.describe().to_string()
    sample = df.head(5).to_string()
    null_counts = df.isnull().sum().to_string()

    response = client.messages.create(
        model="claude-opus-4-6",
        max_tokens=3000,
        system="You are a senior data analyst. Produce executive reports with clear insights and actionable recommendations.",
        messages=[{
            "role": "user",
            "content": f"""Report goal: {report_goal}

Dataset summary:
{summary}

Sample rows:
{sample}

Missing values:
{null_counts}

Generate a report with: Key Findings, Anomalies, Trends, and Recommendations."""
        }]
    )
    return response.content[0].text

Stack: n8n + Google Sheets + Claude API + Google Slides API

  1. Data Pull: n8n fetches data from Google Sheets or a database on a schedule.
  2. Aggregation: Basic computations (totals, averages, top performers) via n8n Code node.
  3. Analysis: Aggregated data sent to Claude for narrative insights and slide-by-slide recommendations.
  4. Presentation: A Google Slides template is populated with the generated content via Slides API.
Back to top