Open-source Deep Research

Deep Research for Claude Code

Give Claude Code a cited research workflow without moving prompts away from your editor.

MCP setup fits Claude Code

AutoSearch connects through Claude Code's MCP flow, so the host keeps planning and tool selection while AutoSearch handles source discovery. You install the server, expose the tools, and keep research prompts inside the same coding session where decisions are made.

Research without model lock-in

Claude Code can keep using its own model while AutoSearch retrieves evidence from 40 channels. That separation is useful when teams want reproducible source coverage, cited notes, and the option to change agent hosts without rebuilding their research workflow.

Chinese sources in coding loops

Many implementation details appear first in Chinese blogs, forums, and product communities. AutoSearch gives Claude Code access to 10+ Chinese sources alongside GitHub, arXiv, Reddit, and Hacker News, making multilingual technical discovery part of normal agent work.

How it fits

AutoSearch sits beside Claude Code as an MCP-native research server. Claude Code remains the place where you ask questions, inspect code, and decide what to change. AutoSearch receives the research task, fans it out across technical, academic, social, and Chinese channels, then returns cited findings that Claude Code can reason over. The workflow is especially useful before migrations, dependency choices, implementation planning, and competitive scans because evidence comes back without coupling the research system to one LLM provider.

Try this prompt

Research current best practices for adding MCP tools to Claude Code.
Compare official docs, GitHub examples, and recent developer discussions, then cite each recommendation.