Researchers: turn "papers I have read" into a real knowledge base
Save PDFs, web articles, and your own notes into the vault, and the AI agent answers from what you have actually read — not from whatever a search engine surfaces.
The agent reasons over your notes, not a random search result.
Who it is for
Scholars, PhDs, product researchers, market analysts — any role where knowledge density matters.
What hurts today
Zotero stores PDFs, Obsidian holds notes, ChatGPT answers questions. Three silos. Asking ChatGPT to synthesize "everything you have read on X" — it cannot see any of it.
How Kition helps
Kition lets the agent read your vault directly. Ask "which DPO papers did I read last year?" and it runs a real search, citing the .md file path. Conclusions are traceable.
Suggested workflow
Step 1
Capture sources
web_article_save converts web pages or arXiv papers to Markdown with the original URL and a summary.
Step 2
Reasoning & synthesis
The agent uses fs_grep and fs_read on your notes; conclusions cite file paths.
Step 3
Project board
A table for the reading list, experiment log, and idea backlog.
Step 4
Collaboration
git push the vault to coauthors — the reasoning trail travels with the notes.
Outcomes
- AI reasons over your domain knowledge, not the internet at large
- Every claim is traceable to a paper or note
- Research thinking is preserved end-to-end
- Local + Git collaboration keeps sensitive data in-house
Ready when you are.
Kition is a local-first AI workspace. Markdown documents, structured tables, and an AI agent — running on your own machine, against the model provider you choose.