Kompas AI positions itself as a tool that works alongside your thinking, with a focus on capturing ideas, connecting them, and developing them into something truly usable. It collects your reading, conversations, and unformed ideas, places them in one place, builds connections, and transforms them into work that is distinctly your own.
Purpose Kompas is for people who want their ideas to keep growing rather than merely be saved. One problem it hopes to solve is this: we read more than ever, talk with AI all day, and produce thousands of words, yet almost nothing stays with us—the information stream slides past, the model answers and we nod and close the tab, nothing accumulates, and so next week we think the same things all over again. Kompas tries to be a correction to this situation: every session lands somewhere you can find again and connects to what you have written before.
Main capabilities According to the materials, Kompas's capabilities include:
- Sessions, notes, and graph: every session leaves traces—decisions, memories, retrieved passages—and connects to what you have written before.
- From fragments to original ideas: documents, highlights, voice notes, and web pages are marked as belonging to you before anything else touches them.
- Start thinking from a topic: starting from a topic, Kompas brings in your past notes, decisions, and reading, thereby asking you better questions in return.
- Sessions become drafts: a session can be moved into a draft, past work appears when useful, and writing begins entirely from your own thinking.
- Writing assistance approach: Kompas does not write for you, but writes with you—by remembering what you have already written, it helps you make good use of it at the moment you need it.
- Graph visualization: sessions form a graph, and the graph becomes your writing.
How to use it and runtime environment Kompas is always a desktop app, not a browser app. The materials state that it is available for macOS (Apple Silicon and Intel) and Windows. The browser is used only for registration, account, and billing pages.
AI access varies by version:
- Free and Pro Sync: bring your own OpenAI or Anthropic API key, use your Claude Code or Codex subscription, or run local models via Ollama. Real-time voice and audio transcription require an OpenAI key.
- Pro Full: built-in frontier AI, provided through Kompas Cloud, with no personal key required. When you activate Pro Full, previously saved personal API keys willDiscontinued; if canceled, it will automatically reactivate.
Version and pricing direction The material gives three tier directions:
- Free: unlimited local sessions, library, writing, and graph; bring your own OpenAI or Anthropic key; can use Claude Code or Codex subscription; can run local models via Ollama; real-time voice and audio transcription require an OpenAI key; single device; export all content as JSON at any time; MIT open-source license.
- Pro Sync: includes what is in Free, plus cross-device sync between desktop devices, and hosted graph storage with a 5 GB soft limit / 10 GB hard limit; still requires bring-your-own key, subscription, or Ollama; does not include Kompas Cloud LLM credits or top-ups; can upgrade to Pro Full at any time.
- Pro Full: includes what is in Free and Pro Sync, plus cross-device sync between Pro desktops; built-in frontier AI, no API key required; includes AI usage, with visible limits and optional top-ups; offers a 7-day free trial, no credit card required; account and billing are handled in the browser, Kompas itself remains a desktop app; priority email support. The material says pricing is "honest and straightforward." Specific price figures are not provided in the material.
Data and privacy The material clearly states:
- In the free desktop version, your sessions, notes, and graph stay on your device; the only network calls are to the LLM provider you choose, subject to its own data policies.
- In the Pro version, the graph is privately hosted by Kompas to enable cross-device sync, but your content is still not used to train models.
- Your content is never sold, never used to train models.
- Under the Free and Pro Sync versions, browsing and editing local data can work offline; the network is used only when calling an LLM or using voice transcription. Ollama can make text-based AI features run offline, while voice transcription still requires an OpenAI key. Pro's built-in AI and cross-device sync require a connection; when offline, local data can still be read and edited, and sync resumes after reconnecting.
Open source and self-hosting The desktop app is open source under the MIT license. Only the cloud code—billing, hosted models, and abuse detection—is in a separate private repository. The hosted Pro cloud service is operated by Kompas, and self-hosting is not offered; the desktop build is the self-managed option, with the same retrieval, sessions, writing, and graph features, fully runon your own machine.
Use cases The material lists the intended audiences as including: people who want ideas to grow rather than merely be saved; people who want a draft to carry their own argument rather than the model's argument; people engaged in ongoing thinking, letting ideas grow and revealing connections between concepts; and people who want to remember past decisions and why they were made. In addition, Obsidian and Logseq users are also mentioned, suitable for moving beyond collecting material toward expanding their own thinking based on the material.
Limitations and missing information The material does not provide specific price amounts, the specific values of AI usage quotas, the list of specific models supported, supported languages, team size, or company information, etc. The detailed policy for Pro data handling is published on the privacy page, but the material does not elaborate on its contents. Any personal API key is deactivated while Pro Full is active and automatically reactivated after cancellation. Pro features remain available until the end of the current paid period, after which the desktop version falls back to Free mode: saved personal API keys are reactivated, sync stops, local data remains unchanged; the hosted graph is retained for a 30-day grace period, during which resubscribing can restore sync; it can be exported as JSON at any time before expiration.