research-hub plugin
A local researcher workspace for papers, evidence and drafts in one project, with AI proposals that people review and approve.
- 67 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The workspace reads and registers project files and hashes, and a connected Codex receives manuscript text excerpts.
- Source installation puts dependencies into a virtual environment.
Install
- Works in
Manual install
git clone -c core.autocrlf=false https://github.com/WenyuChiou/research-hub.gitThen create a virtual environment and run `pip install -e .`. The exact revision and commands are in the README.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The workspace reads and registers project files and hashes, and a connected Codex receives manuscript text excerpts.
- Source installation puts dependencies into a virtual environment.
This is third-party code. Review the repository files before installing.
What it does
research-hub brings questions, sources, evidence, results and manuscripts into one project on your computer. The workspace registers file paths and fingerprints (hashes) to notice input changes and return related sections for review. Pages cover overview, literature, evidence linked to sources, design and results, the manuscript (Word, LaTeX or Markdown) and a review stage with local delivery. AI proposals need human review and local delivery is approved separately. For literature discovery in an agent the portable research-hub skill is used with working search and reading tools; installing Hub, Zotero, Obsidian or NotebookLM is not required. MCP and REST integration connect agents, and the Claude plugin ships `.claude-plugin/plugin.json` and `SKILL.md`.
Who it is for. Researchers and graduate students who manage literature, evidence and manuscripts with an AI agent and want decisions to stay with them.
Good fit when
- Sources on a topic are needed with original links, versions and real reading limits.
- Every manuscript claim must be tied to a source and unsupported spots made visible.
- A results file changed and related sections must return for review.
- AI drafts a text revision and the author accepts or rejects it after review.
Not a fit when
- You need automatic search and full-text reading with no human: search snippets do not count as reading.
- You need a ready cloud service; the workspace is local.
- You want scientific conclusions checked for you; a resolvable link does not prove it supports a claim.
Example request
Find papers on LLMs simulating human decisions with original links, versions and limits of available full text.Limitations
The workspace needs Python 3.10 or newer, Git and a local browser. The source version is separate from the `research-hub-pipeline` PyPI package, which holds the older literature dashboard and not the new workspace. MCP needs the optional `[mcp]` extra. A resolvable identifier does not prove a source supports a claim, so check scientific judgments yourself. A connected Codex receives only bounded text excerpts.
How to disable. Delete the cloned folder and virtual environment. Workspace directories such as `workspace-demo` and `.research_hub` holding your data are removed separately.
FAQ
Do I need Hub to search literature?
No. An agent with working search and reading tools plus the portable skill from `skills/research-hub/` is enough. Hub, Zotero, Obsidian and NotebookLM are not required.
How do I launch the workspace?
Clone the repository, create a virtual environment, run `pip install -e .` and follow the demo launch guide in the README.
How does the skill separate searching from reading?
It requires stating real reading limits. Search snippets do not count as full-text reading and unavailable sources are marked unknown.
Does the AI decide automatically?
No. Proposals require human review and local delivery is approved separately. Nothing is published automatically.
What is the research-hub-pipeline PyPI package?
It is the older literature dashboard. The new workspace installs from source, so the package version does not signal the new features.
Discussion
No questions or reviews yet.
Related
PluginsA marketplace of five Claude Code plugins for researchers: literature search, research design, project memory and manuscript work
A research loop that keeps evidence: a literature review, paper review or survey runs in steps with checks and resume.
MCP server with 40 tools that checks quotes, numbers, DOIs and PRISMA counts in AI-written literature reviews. No model inside.
Research Weave turns a pile of papers into a reading queue, an evidence map and a sharper research question
Research Skills gives Claude Code, Codex and Copilot CLI plugins for research: literature, grants, papers, figures, slides.
cli-anything-zotero: Zotero for AI agents
cli-anything-zotero
The zotero-cli command manages a local Zotero library: search, DOI import, BibTeX export, PDFs, notes, DOCX citations
README in short
The repository offers two paths: use the skill in an existing agent or run the workspace with a demo on two seed references. Detailed docs cover architecture, MCP tools and file formats.