An open-source deep research assistant that checks its citations and keeps what it found.

What it is for

Three people, one problem.

Your vet mentions a new drug for your dog's cancer. The follow-up is in three days. You want to know what the studies say: how well it works, for which breeds, and what the side effects are. The answer is spread across twenty papers, a regulator's review, and many pages that only look like medical advice.

You are tracing your family. Your great-grandfather is in the 1911 census in Osaka. By 1921 he is gone from every record the family knows about. Somewhere there are ship manifests, a prefecture's emigration lists, and an old newspaper notice. Most of them are in Japanese. None of them are on the first page of a search.

You are writing one chapter of a thesis on the influence of the Soviet-Afghan War on 1990s British science fiction. You have forty papers and articles saved, in three languages. You have one week to work out what they agree on, where they disagree, and what nobody has looked at yet.

None of that can be read in one evening. That is the job ResearchZosho does. You give it the question. It reads and notes each fact with the source it came from. When it is done, a write-up is waiting: what it found, where the sources disagree, and what it could not resolve. Every claim in it points at a source you can check. It keeps what it found, so the next question on the same subject does not start from zero. If a paper you used gets a new version, or two of your own findings contradict each other, it tells you.

Nothing counts as valid until you have reviewed it.

It runs on your own computer with a model you pick, local or hosted. It is free and open source. It is for anyone who needs reliable research done.

This is The Research Harness For The Rest Of Us.

What you can do with it

• Talk to the Librarian. researchzosho chat, or the Chat page. Ask what the library holds, ask a follow-up, say "find out" to start a research run, or decide the inbox. It answers only from what it looked up, and says so.
• Send it a question. It reads in parallel, in more than one language when the question calls for it. It writes the answer up in sections, for as long as you allow.
• Check the answer. Every write-up ends with a table of the facts it used, each with a source and a quote. Each cited sentence is checked against its source before it is shelved.
• Keep bad sources out. One source is a draft until a second, independent one backs it. A retracted paper disputes what cites it. Sites you ban are not used. Sites you trust count as primary.
• Decide what stands. Accept, dispute or retire each claim. The library uses what you accept next time.
• Bring your own documents. A PDF, a web page or a whole folder, shelved as a collection, kept or read in place. A chat you had with another assistant, a reading list, your bookmarks, a meeting transcript, or your own draft: each becomes shelves, open questions and claims to check.
• Start from a repository, a paper or a page. researchzosho survey reads a code repository, a paper, a website or an issue tracker. It says what it is and what it rests on, and lists research questions about it. Pick some, or write your own.
• Ask your own data. researchzosho db add gives the library read-only access to a database: SQLite, PostgreSQL, MySQL, SQL Server, MongoDB, or a folder of CSV files. Research runs and the chat can query it, and a report cites the query it ran.
• Get book recommendations. Which book first on a subject, what a book covers, what to read next. The recommendations come from the library catalogues, never from the model's memory, and say what each book covers and who it is for.
• Use your own shelf. Point it at your Calibre library. It learns what you go for, recommends books on your shelf like one you name, and the lists move with what you liked and disliked.
• Know what you own. Import your purchase records and it can tell what you already have. Every night it checks the product-safety feeds against them, and a recall that may concern you waits in the inbox.
• Research a purchase. What shops list right now in your country, with the price, the datasheet, recalls, repairability and what owners say. Fit against your need, not popularity.
• Ask how to do or make something. A craft, a hobby, a sport, a martial art or a trade: the how, and the kit with it, every material and tool, what stands in for it, and where to get it. For a community you want to join, how to join and what its rules ask of your equipment.
• Connect two subjects. researchzosho bridges finds two subjects in your library that share a concept but that no source connects. It writes a research question about them. You run it like any other. It also says how far apart any two subjects are.
• See how things connect. The map draws the people, places and works around a name.
• Open it in a browser. Search, ask, read a claim with its sources, walk the subjects, see what changed, send a question, download a write-up as Markdown or PDF. Or open the vault folder in Obsidian or SoloMD.
• Let other programs use it. Claude Code, Codex, Gemini CLI and any other MCP host, plus Python and Java clients.
• Reach it from outside your home. researchzosho share makes the library reachable in half a minute, free, with no port opened at home. Anyone with the link sees only the sign-in page.
• Sign in simply. A name and a password, from any phone or computer; a passkey or a one-time code if you want them. Invite a person with one link.

The guide has each of these in full.

Install

Current release: 0.6.5. Every artifact, with checksums, is on the download page.

Linux and macOS:

curl -fsSL https://researchzosho.org/install | sh

Windows, from PowerShell:

irm https://researchzosho.org/install.ps1 | iex

Both scripts download the release from GitHub and check it against that release's SHA256SUMS before installing. If the hash does not match, they stop. Only the script comes from this domain. The program and the checksums both come from the GitHub release. If you would rather not pipe a script into a shell, download it and read it first. It is short.

Research runs need a model server that speaks the OpenAI chat API: a local one (llama.cpp, Ollama, LM Studio) or a hosted API with a key (OpenAI, DeepSeek, Gemini, OpenRouter and others). You also need Java 21 or newer, unless you take a build that carries its own runtime. The installer picks that build when Java is missing. ResearchZosho ships neither. Your documents go only to the model server you name. Asking what the library already has works without a model.

Which model to run, by VRAM (the graphics card's memory, not the computer's RAM), is on the models page: the measured list, from the 27B down to what fits in 4 GB.

Research runs also need a web search backend. It matters as much as the model. Use a Brave Search API key (there is a free plan), or SearXNG, which setup can start with Docker. With neither, the built-in fallback searches Wikipedia and the scholarly literature only.

Then:

researchzosho setup

setup asks a few questions, each with an answer already filled in. Then it shelves a document you name and answers a question about it. When it is done, open http://127.0.0.1:4649/ in a browser. If you already use CodeZaiku, codezaiku install researchzosho does the download, the check and the setup in one go.

Documentation

• README: the whole project on one page.
• LIBRARIAN_HOWTOUSE.md: how to use it day to day.
• LIBRARY_PROTOCOL.md: for programmers, how a program talks to it. Sections 1 to 6 are the contract and keep working across releases.