Commons

A local-first research system for doctoral students. Bring your own model.

A doctorate asks you to read more than you can hold in your head, to keep track of what every source actually says, and to test your argument against the evidence that cuts against it. General-purpose chatbots make that harder: they paraphrase loosely, invent citations, and cannot be trusted with interview data. Commons runs on your own computer, with whichever models you choose, and keeps every claim tied to a verbatim quote and a page.

What it does

• A knowledge base you ownEvery source becomes a plain Markdown note with its verified claims. The Markdown is the source of truth.
• Claims checked mechanicallyA claim enters the record only if its quote appears word for word in the source. A second model from a different family then judges it. Doubt goes to you, not into your evidence.
• Your thesis, testedWrite your propositions and the rival explanations an examiner would offer. Commons shows the strongest evidence each way, and looks for evidence against you as hard as evidence for you.
• The literature, every morningA daily sweep of OpenAlex, Crossref, arXiv and your own feeds, screened against your propositions, with a digest waiting at breakfast.
• Answers from your corpus onlyNumbered citations, checked against the passages, or a plain statement that your corpus does not hold the answer. Word reports with real charts, Excel, BibTeX and RIS.
• Interview data stays homeMaterial under ethics approval only ever reaches a model on your own machine, or an institutional service your approval names. Never a cloud API.

Start

pip install "phd-commons @ git+https://github.com/phd-commons/commons"
commons init ~/Commons --preset ollama --title "My thesis"
cd ~/Commons
commons doctor

Then open a coding agent in that folder and run /commons-setup. It reads your proposal, drafts your propositions and rivals, picks models that fit your machine, sets up the daily feed and seeds your corpus from your bibliography.

Models

Ollama, LM Studio, MLX on Apple Silicon, llama.cpp, vLLM, a university AI gateway, or a hosted API. Every provider is labelled local, institutional or cloud, and Commons enforces what each may see.

Source and citation

github.com/phd-commons/commons · Apache 2.0 code, CC BY 4.0 documentation · cite with the repository's CITATION.cff.