In one sentence
Kition is an AI workspace for documents, structured data, research, and automation. Markdown documents, data tables, browser research, visual workflows, and Agent sessions share one project context instead of living in disconnected tools.
The shortest useful comparison is: an Obsidian-style editor, Notion-style tables, and a tool-using AI agent designed as one product. The point is not to clone three applications. The point is to remove the handoffs between writing, structured data, and automation.
The problem Kition is trying to solve
Knowledge work is usually split across incompatible surfaces. Notes live in one application, project trackers in another, source files in a folder, and AI in a separate browser tab. Every transition loses context: you paste text into chat, explain where it came from, copy the answer back, and then manually update the table that tracks the work.
Kition treats the vault as the shared source of truth. A document can explain a project, a table can track its records, and the agent can read or update both with your permission. This is why the agent is built into the workspace rather than added as a floating assistant after the rest of the product was designed.
Three primitives, one vault
The product is easier to understand as three primitives that deliberately overlap. They use the same local vault, so links and automation do not have to cross a proprietary integration boundary.
- Documents: a CodeMirror-based Markdown editor with live preview, wikilinks, backlinks, callouts, code fences, math, diagrams, and a command palette. The saved file is still readable in VS Code, Vim, Obsidian, or any other Markdown editor.
- Tables: structured records with field types and Grid, Kanban, Gallery, and Calendar views. Tables can be embedded in documents, imported from CSV, exported again, and accessed by the agent as structured data rather than a screenshot.
- Agent: tools for files, commands, documents, the web, browser sessions, tables, and MCP servers. Plan mode, permissions, hooks, and subagents exist to make longer workflows observable instead of hiding them behind a single prompt.
How workspace ownership works in practice
You choose a vault directory. Documents are stored as .md; tables use .kitable; workspace metadata lives under .kition/. You can inspect the directory, back it up, put it under Git, sync it with a tool you trust, or leave Kition without exporting every note through an API.
Kition also follows a bring-your-own-model approach. OpenAI Responses, Anthropic Messages, and compatible endpoints can be configured side by side. Provider credentials are stored through the operating system credential store rather than written into the vault. The detailed security model is covered in Securing your API key.
What the agent can and cannot do
A Kition agent can inspect files, propose document edits, query tables, fetch web pages, use a browser session, run permitted commands, and call tools exposed by mounted MCP servers. That makes it suitable for tasks such as turning research into notes, maintaining a content calendar, triaging issues, or checking a folder before producing a report.
It is not supposed to operate without boundaries. Tool permissions can allow, ask, or deny actions; destructive work should require confirmation; and hooks can record or block tool calls. The model still makes mistakes, so important changes need review. Kition exposes tool activity and diffs because an agent that can act should also make its actions inspectable.
A realistic first session
- Install the desktop app from the download page and create a new empty vault.
- Create one Markdown document describing a project, decision, or research topic you already understand.
- Configure a provider using the provider setup guide, then run the built-in connectivity test.
- Ask the agent to summarize the document and propose a clearer outline. Review the diff instead of accepting blindly.
- Create a small table with status, owner, and due-date fields, then ask the agent to add two sample records.
- Close Kition and open the vault in a normal file browser or editor. Verify that the documents remain ordinary files you control.
Who Kition is a good fit for
Kition is aimed at people who want their working context close to the filesystem: developers, researchers, writers, consultants, product teams, and serious personal-knowledge-management users. It is especially relevant when local ownership, model choice, or agent extensibility matters more than instant browser-based collaboration.
It may not be the right choice when your primary requirement is a mature cloud collaboration suite with hundreds of integrations and zero local setup. A team that lives entirely in Notion or Google Workspace may prefer those collaboration models. Kition is strongest when the vault itself should remain portable and the agent needs controlled access to real working files.
How to evaluate it without buying into the story
- Open the vault outside Kition and confirm the documents are readable.
- Export a table to CSV and confirm the resulting data is usable elsewhere.
- Disconnect the network and check which non-provider workflows still function.
- Inspect where provider credentials are stored and verify they are absent from the vault.
- Run a small agent task and review every tool call and proposed edit.
- Compare the workflow with your current stack using the same real task, not a demo prompt.
Where to go next
For the complete product model, read What is Kition. Then follow Install Kition, Create your first vault, and Configure an AI provider. The full feature list is useful when you need to verify a specific capability before installing.