Open-source deep research agents: how they work and how to use them

Exa’s guide explains how deep research agents generate structured, cited reports by combining language models, web search, page retrieval, and stopping rules, contrasting configurable open-source systems with simpler managed products. It uses LangChain’s Open Deep Research as an example of an agent that can run locally or through LangGraph, supports multiple models and search tools, and can connect to Exa through MCP for web search and page highlights; users remain responsible for API keys, infrastructure, costs, and research limits. Commercial offerings from ChatGPT, Gemini, and Claude require less setup but restrict model and search-provider choice, while providing cited research reports within plan-based usage limits. The guide also describes fully local configurations using Ollama, LMStudio, DuckDuckGo, and optionally self-hosted SearXNG for greater privacy, noting that web searches may otherwise still leave the user’s device. For teams seeking managed infrastructure, Exa’s Agent API can conduct multi-step search, reading, and verification through a single endpoint, returning structured JSON with citations and predictable effort-based pricing.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.