Research agent
A research agent is an LLM-driven agent specialized for multi-source synthesis — searches, reads, summarizes, compares, cites — packaged either as a product feature ([[deep-research-mode]]) or as a custom agent in n8n / Mastra / LangGraph / Pipedream.
Research agents differ from chat assistants in three ways: longer time-budget (minutes vs seconds), multi-source synthesis (10-50 pages vs 1-3), and structured-report output (sections + citations vs single answer). Production patterns: (1) consumer products embed Deep Research as a button (ChatGPT, Perplexity, Gemini, Claude), (2) B2B teams build custom research agents on top of frontier APIs + a search tool ([[web-retrieval-tool]]) for domain-specific workflows (M&A diligence, competitive intel, regulatory monitoring), (3) open-source frameworks (GPT Researcher, OpenDeepResearch) ship as starting points. The bar is quality + citation honesty; speed is secondary.
When to use research agent
- Domain-specific multi-source synthesis where chat is too shallow.
- Regulated workflows where citations + evidence trails matter.
Common mistakes
- Skipping per-source quality filtering — agent eats SEO-spam citations.
- Not capturing the trace — hard to debug when the agent reaches a wrong conclusion.
FAQ
What is research agent?
A research agent is an LLM-driven agent specialized for multi-source synthesis — searches, reads, summarizes, compares, cites — packaged either as a product feature ([[deep-research-mode]]) or as a custom agent in n8n / Mastra / LangGraph / Pipedream.
When should I use research agent?
Domain-specific multi-source synthesis where chat is too shallow. Regulated workflows where citations + evidence trails matter.
What are the most common mistakes with research agent?
Skipping per-source quality filtering — agent eats SEO-spam citations. Not capturing the trace — hard to debug when the agent reaches a wrong conclusion.
Related terms
- Deep research mode — Deep research mode is the multi-hour LLM-agent feature that runs an autonomous research loop — searches dozens of sources, reads, synthesizes, cites — and returns a structured report. OpenAI Deep Research, Perplexity Pro Research, Gemini Deep Research, Claude Research are 2026 examples.
- Search-grounded generation — Search-grounded generation is the LLM workflow where every output sentence is anchored to a retrieved source — the production pattern behind Perplexity, ChatGPT Search, Gemini AI Overview, You.com.
- Agent loop — An agent loop is the repeating cycle of an AI agent — observe state, decide on an action (usually a tool call), execute, observe the result, and repeat — until a goal is reached or a stop condition fires.
Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/research-agent.md.