Build a Multi-Agent Research Assistant
Chain together multiple AI models, each with a specialized role, to conduct deep research on any topic, fact-check findings, and produce a structured report automatically.
Download workflow config (JSON)How the data moves
This is a handoff plan, not an automatic integration. Review outputs before the next step. Unlabeled arrows show order, not a verified data format. Use the capability links to explore tool substitutions.
Step 1
Design Your Agent Pipeline
Map out your agent roles: (1) Query Planner, breaks the research question into sub-queries; (2) Search Agent, fetches data from the web; (3) Synthesizer, combines findings; (4) Fact-Checker, validates claims. Use Claude to help design this architecture.
Step 2
Set Up the Query Planner Agent
Create a Make scenario where Agent 1 (ChatGPT) receives a research question and outputs 5–8 specific sub-queries needed to answer it comprehensively.
Step 3
Deploy the Search Agent
For each sub-query, trigger a Perplexity search via Make's HTTP module. Perplexity's API returns answers with citations, store each result with its source URL.
Step 4
Run the Synthesizer Agent
Pass all the gathered sub-answers into Claude with the original question. Claude synthesizes findings into a structured report: Executive Summary, Findings by Section, Key Contradictions, and Source List.
Step 5
Fact-Check Critical Claims
Extract factual claims from the report and run them through a second Perplexity pass to verify. Claude cross-references the verification results and flags any claims that couldn't be confirmed.
Step 6
Format and Deliver
Use Make to send the final report to Notion AI for formatting and tagging. Trigger an email notification with a link to the new report page.
Make It Yours
The speed is real. Here's what keeps the output from reading like everyone else's.
- •Write the Query Planner's instructions around the kind of research you actually do repeatedly, not a generic "research anything" brief.
- •Give the Fact-Check agent your own bar for source quality (primary sources only, or a minimum publication date) instead of the model's default judgment.
- •Set the report format to match how you actually consume research, a bullet brief versus a full memo, so the output gets used instead of skimmed once and shelved.
Time Required
4 hrs setup / 20 min per report
Expected Result
A structured, fact-checked research report on any topic, produced in 20 minutes by a chain of specialized AI agents rather than hours of manual research.
Related Workflows
Build a Multi-Agent Research Pipeline with CrewAI
Set up a CrewAI pipeline where specialized agents handle different research tasks in parallel, one searches papers, one synthesizes findings, one checks contradictions, delivering a comprehensive brief automatically.
View workflowBuild a Price and Competitor Monitoring System
Stay automatically updated on competitor pricing, product changes, and positioning shifts, without manual checking or paying for expensive intelligence tools.
View workflowRun a Multi-Agent System Like an Enterprise AI Engineering Team
Get a coordinated multi-agent system running on a real task in about a day.
View workflow