Kompas AI
Kompas is a desktop chat and knowledge app for macOS and Windows that turns your reading, conversations, and half-formed ideas into a connected graph you can search and write from. You capture documents, highlights, voice notes, and web pages locally, then run sessions where Kompas pulls your past notes and decisions into the thread. Sessions leave traces linked to earlier work, so today's thought can brush against something you wrote last month. The free desktop build keeps everything on your device; Pro adds cross-device sync and optional managed AI.
Most chatbots answer and forget. Kompas is built for people who want conversations to compound: it remembers what you already wrote and surfaces it when it matters, without training models on your content. The trade-off is scope: there is no web app, only a desktop binary, and Pro cloud sync is not self-hostable. If you want a persistent thinking workspace rather than a one-off Q&A box, that narrow focus is the point.
Essayists use Kompas for drafts that carry their argument, not the model's voice. Researchers keep continuous note chains and spot links between concepts. Founders revisit decisions with the reasons they recorded earlier. Obsidian and Logseq users who tired of only collecting material can extend thinking inside the same graph-backed sessions.
Free desktop app for macOS Apple Silicon, Intel, and Windows x64 with unlimited local sessions and graph storage
Pro Sync at $3 per month adds cross-device sync with 5 GB soft and 10 GB hard hosted graph storage
Pro Full at $38 per month includes frontier AI, a 7-day free trial, and no credit card required to start
Bring your own OpenAI, Anthropic, Claude Code, Codex, or Ollama keys on Free and Pro Sync tiers
Desktop app is open source under the MIT license at github.com/bc-dunia/kompas
Kompas does not train on your content; Free keeps data local and Pro hosts graphs privately for sync only
Sessions link to past notes, decisions, and reading so thinking compounds instead of resetting each chat
Free tier keeps your graph on your device with no training on your content
Flexible AI: bring your own keys, Claude Code or Codex subscription, Ollama, or Pro Full managed AI
Open source MIT desktop build with JSON export on every tier
Pro pricing shows included AI usage clearly with optional top-ups and no surprise charges
No web app; macOS and Windows desktop only
Pro cloud service is not available for self-hosting
Realtime voice and audio transcription on Free require an OpenAI API key
Does Kompas train AI models on my sessions?
Kompas does not train on your content. On the free desktop app, sessions, notes, and your graph stay on your device. On Pro, Kompas hosts your graph privately for cross-device sync and still does not use your data for model training.
How does Kompas differ from ChatGPT or Claude?
ChatGPT and Claude are chat interfaces you talk to and close. Kompas is a desktop workspace where conversations, documents, and half-finished arguments live together and surface when today's thought connects to last month's notes.
Which AI models can I use with Kompas?
On Free and Pro Sync, Kompas lets you bring your own OpenAI or Anthropic API key, Claude Code or Codex subscription, or local models via Ollama. On Pro Full, Kompas includes frontier AI through Kompas Cloud with no personal API key required.
Is Kompas open source?
Yes. The Kompas desktop application is open source under the MIT license on GitHub. Cloud-only components such as billing, hosted models, and abuse detection live in a separate private repository.
What happens if I cancel my Kompas Pro subscription?
Kompas Pro features keep working until the end of your paid period, then the desktop falls back to Free mode. Your local data stays intact, sync stops, and your hosted graph is kept for a 30-day grace period before purge.
Does Kompas have a web app?
No. Kompas is always a desktop app for macOS and Windows. The browser is only used for Pro sign-up, account management, and billing pages at app.kompas.ai.
Can I use Kompas offline?
On Free, Kompas lets you browse and edit local data offline; network calls happen only when you invoke an LLM or voice transcription. On Pro, built-in AI and sync need a connection, though you can still read and edit locally while offline.

