HyperResearch: Agent-Driven Research Knowledge Base

TL;DR

TL;DR: HyperResearch is an open source research knowledge base built for AI agents. The agents collect web research, organize it into a persistent corpus, and synthesize answers from it. It is the project to reach for when you want a research surface that survives across sessions, rather than a one-off deep-research query that disappears.

What Is HyperResearch?

The repository description: “Agent-driven research knowledge base. Agents collect, search, and synthesize web research into a persistent, searchable [corpus].” HyperResearch targets a specific failure mode of one-shot deep-research tools — they run, they return, and the next session has to redo the work.

The topics — agents, agentskills, claude-code, deep-research — confirm the integration story. HyperResearch is designed to plug into Claude Code as skills, so the agent loop you already use can write to and read from a long-lived research corpus.

How It Works

The Python codebase ships an agent-driven research loop. The implied pipeline:

• An agent reads a research question.
• It runs web searches, fetches pages, and pulls structured data.
• It writes findings into a persistent knowledge base — local, queryable, versioned.
• It synthesizes answers from the corpus rather than redoing the search every session.

The corpus is the differentiator. Most deep-research tools end at “here is your answer.” HyperResearch ends at “here is your answer, and here is the corpus you can keep querying and extending.”

Repo Facts

• Repository: jordan-gibbs/hyperresearch
• Language: Python
• License: MIT
• Stars at time of writing: 2,639
• Topics: agents, agentskills, claude-code, deep-research

When To Use It

HyperResearch is the right tool when:

• You do recurring research on a topic and want the corpus to persist across sessions.
• You use Claude Code or a similar agentic CLI that supports skills.
• You want the agent loop to own the research workflow, not a one-off deep-research call.

It is not a search engine, it is not a citation manager, and it does not replace reading the primary sources yourself. It is the persistence layer between your research and the next time you need it.

Source and Accuracy Notes

• Project page: https://github.com/jordan-gibbs/hyperresearch
• Source repository: jordan-gibbs/hyperresearch (verified via GitHub API html_url)
• License: MIT (verified via GitHub API license.spdx_id)
• Stars at time of writing: 2,639
• Deep-research and Claude Code are referenced as integrations; verify the project’s current skill format and install instructions before relying on them in production.