简体中文
English
Website review
This review may include AI-generated content and is not individually verified. Domain, certificate and other facts come from the observed data on this page.
What is Kompas?
Kompas is a desktop thinking tool that captures your reading, conversations and half-formed ideas in one place, links them together as a graph, and helps you turn them into writing. Its pitch is that ideas should compound rather than be re-thought from scratch: sessions are stored so you can find them again and see how they connect to earlier notes, and a Writer view carries that graph into drafts. The free version is local-first and open source (MIT), letting you bring your own OpenAI or Anthropic key, a Claude Code or Codex subscription, or local models via Ollama. A Pro cloud tier adds cross-device sync and built-in AI with included usage limits and optional top-ups, offered with a seven-day trial.
Who it suits, and the trade-off
It fits people who read and talk to AI constantly but feel nothing accumulates—researchers, essayists, students building an argument over weeks. The trade-off is setup and control versus convenience: the free desktop route keeps ideas on your device and gives you model choice, but you manage keys and models yourself; Pro removes that friction and adds sync, at the cost of your material living in the cloud and AI usage caps.
A concrete test: take a note you wrote a month ago and one from today. If you can't see a line between them, the graph-and-session model is worth trying; if you mainly need a fast scratchpad, it may be more structure than you need.
As a next step, run one real project through it—capture a few sources, hold a session, then draft in the Writer—and judge whether the connections it surfaces are ones you'd have missed. For a broader note-taking alternative, see Obsidian.
How does Kompas turn my scattered notes and conversations into a connected graph I can write from?
Kompas works by making every session land in a persistent, linked store rather than disappearing when you close a tab. The page describes a three-step flow: you capture fragments (documents, conversations, half-formed ideas), the tool connects them into a graph of related sessions and notes, and then a Writer view lets you draw on that graph to produce your own writing. The graph is the middle layer—not a separate chore, but a byproduct of sessions you were already having.
What the capture-to-graph loop looks like
The page's session example shows the mechanics concretely. Within a session on patience, Kompas surfaces an earlier session ("in session 009, you wrote that retrieval is the act of remembering you already chose") and a saved note quoting Camus. A side panel lists three references—a prior session, a note, a decision—and a small live session map links concepts like Patience, Camus, Sisyphus, waiting and recall. So the graph isn't something you build by hand; it accumulates as sessions reference and echo each other.
Where the writing comes from
The "Writer" appears as a top-level mode alongside Capture, Sessions, Projects, Library and Graph. The claim is that "your graph becomes your writing"—you write from the connected material rather than starting blank. Treat that as the intended workflow rather than a guarantee of output quality: the value depends on how deliberately you capture and tag fragments.
Practical trade-offs
| Free (desktop) | Pro (cloud) | |
|---|---|---|
| Where data lives | On your device, local-first | Synced across your desktops |
| AI | Bring your own key, Claude Code/Codex subscription, or local models via Ollama | Built-in frontier AI, no setup |
| Setup effort | Higher—you configure models | Minimal |
| Cost model | Free, open source (MIT) | 7-day free trial, then paid with usage limits and optional top-ups |
The page also states ideas are never sold or used to train models.
A concrete reader scenario
Suppose you read an article, discuss it with an AI assistant, and jot a stray thought. On the free desktop version, all three can be captured locally and linked; when you later open the Writer, the earlier fragments are retrievable as references. If you work across two machines, that local-only setup becomes friction—which is the point at which Pro's sync and built-in AI remove setup work.
Next step: start with a single recurring topic you already think about weekly, capture three or four fragments on it, and check whether the graph surfaces useful connections before committing to either tier. For comparison with other graph-based note tools, see Obsidian and Roam Research.
Should I use the free desktop version with my own API key or pay for Pro cloud sync?
Choose the free desktop version if your ideas should stay on your machine and you already have an OpenAI or Anthropic key, a Claude Code or Codex subscription, or local models running through Ollama. Choose Pro if you work across more than one desktop and would rather not manage keys or model setup at all. The real decision is control versus convenience, not features.
What each path suits
Free desktop (local-first, open source under MIT)
- You are comfortable supplying your own AI access and occasionally troubleshooting it.
- Privacy matters: the page states your ideas stay on your device.
- You want to avoid a subscription and are fine with one machine holding your work.
Pro cloud
- You switch between desktops and want sessions, notes and the graph available everywhere.
- You want built-in frontier AI with no keys and no setup, plus included usage with clear limits and optional top-ups.
- You would rather pay than maintain configuration.
Trade-offs at a glance
| Free desktop | Pro cloud | |
|---|---|---|
| AI access | Your own key, subscription or local model | Built in, no setup |
| Where ideas live | On your device | Synced across desktops |
| Effort | Some setup and upkeep | Minimal |
| Cost model | No subscription | Paid, after a 7-day trial |
A concrete scenario
Suppose you capture fragments on a laptop during the day and want to continue on a desktop machine at night. On the free version, that means moving files yourself or accepting that the graph lives in one place. Pro removes that friction, and the built-in AI means a session can resume with the same context on either machine.
A different reader—someone with a local model already running and material they would rather not send anywhere—gets more from the free version, even if setup takes an afternoon.
How to decide
Start free, as the page itself suggests, and move to Pro when sync and managed AI become the thing you keep wanting. If cross-device work is already your daily pattern, skip the experiment and go straight to Pro; the trial requires no credit card, so you can test whether the sync and built-in AI actually fit before committing.
How does Kompas keep my notes private and prevent them from being used to train AI models?
Kompas's privacy story rests on where your notes live and which AI, if any, touches them. On the free desktop version, the page states that ideas stay on your device and that you bring your own OpenAI or Anthropic key, a Claude Code or Codex subscription, or run local models through Ollama. The desktop app is open source under MIT, so the data handling can be inspected rather than taken on trust. The page also says notes are "never sold, never used to train models."
What that means in practice
- Local-first storage: Your sessions, graph and writing are kept on your machine in the free desktop setup. You control backups and deletion.
- Your own AI account: When you supply an API key or subscription, your notes go to that provider under your account and its terms, not to Kompas as a middleman. Local models via Ollama keep everything on-device.
- Open source: The MIT licence lets a technically minded user or their IT team review what the app sends and when.
- Pro cloud: The page describes cross-device sync and built-in frontier AI with no keys, which necessarily means notes travel to Kompas's cloud. The no-training promise applies here, but the local-only guarantee does not.
Choosing between them
| Need | Free desktop | Pro cloud |
|---|---|---|
| Notes stay on your device | Yes | No, synced to cloud |
| AI provider | Yours (OpenAI, Anthropic, Ollama) | Built in |
| Code you can audit | Yes, MIT | Not stated |
| Multiple desktops | No sync described | Yes |
If confidentiality is the priority, the free desktop version with local models is the most contained option: nothing leaves the machine. If you use a hosted API key, treat your notes as shared with that provider. Pro suits people who want sync and convenience and are willing to rely on Kompas's stated no-training policy.
A practical next step
Write down which of your notes you would be uncomfortable sending to a third-party AI provider. Run those in the free desktop app with Ollama, and keep routine capture in whichever mode you prefer. Before committing to Pro, check the current privacy terms on Kompas for retention periods and how synced data is stored.
Can I use Kompas with a local model like Ollama or my existing Claude subscription instead of buying built-in AI?
Yes. Kompas's free desktop edition is explicitly designed for that: it is local-first, open source under MIT, and lets you bring your own AI. The page names three routes: your own OpenAI or Anthropic API key, a Claude Code or Codex subscription you already have, or local models run through Ollama. The built-in frontier AI is a Pro cloud feature, so you do not need Pro to think with a model you already pay for or host yourself.
What you give up by staying on the free desktop route
- Sync: ideas stay on your device; cross-device sync is a Pro cloud feature.
- Setup: you manage the key, subscription or local model yourself rather than getting AI built in with no configuration.
- Model quality and speed: with Ollama, results depend on the model and hardware you run. A small local model may handle capture, connection and short prompts well but struggle with longer synthesis. An existing Claude subscription avoids that trade-off if it is one of the supported routes.
A practical scenario
Say you read papers and keep half-formed notes, and you already run Ollama on a laptop with a capable GPU. Install the free desktop build, point it at your local model, and run a session on a topic you have circled before. Kompas keeps each session in a graph and connects it to earlier notes, so the value comes from retrieval across sessions rather than from the model alone. If you later want the same graph on a second machine, or want to stop managing keys, that is the point to consider Pro rather than a reason to start there.
How to decide
| Your situation | Sensible starting point |
|---|---|
| You have an Anthropic or OpenAI key, or a Claude Code/Codex subscription | Free desktop, bring your own AI |
| You want ideas to stay on-device and run local models | Free desktop with Ollama |
| You work across several desktops and want AI with no setup | Pro cloud, after trying the 7-day trial |
| You want to inspect or modify the tool | Free desktop, MIT-licensed |
Next step: install the free macOS or Windows build, connect your chosen model, and run two or three sessions on related material. If the graph already helps you resurface earlier thinking, the free route is doing its job; if the friction is sync or key management, that is the specific gap Pro fills. For a broader view of local-first note tools, see Obsidian and Logseq.
How does the session-to-graph-to-writing workflow help me avoid rethinking the same ideas from scratch?
Kompas treats each thinking session as a durable artifact rather than a chat that disappears. The page describes a three-stage loop: sessions you run are captured and linked into a graph, the graph surfaces prior notes and references, and the Writer view turns that accumulated material into your own draft. The point is that a half-formed thought stays findable and connected, so the next session starts from what you already worked out instead of from a blank page.
What each stage does for you
- Session: A working conversation where Kompas can point back to earlier sessions and notes. In the page's example, it reminds you that you "circled this before — in session 009," which is exactly the moment where rethinking from scratch would otherwise begin.
- Graph: Your sessions, notes, decisions and references become a connected map. When you return to a topic, related fragments appear alongside it, so recall replaces reconstruction.
- Writer: The graph becomes source material for writing. You assemble a draft from ideas you already developed, which keeps the output recognizably yours rather than a fresh model-generated take.
A concrete use
Suppose you keep returning to the same question about patience and deliberation. Without a system, each return costs you the same groundwork. With this workflow, the earlier session, the Camus note and the decision you recorded are already linked to the topic. You open the thread, see your past reasoning, and spend the session pushing it forward — testing the reframe rather than rediscovering it.
Trade-offs to weigh
| Free · Desktop | Pro · Cloud | |
|---|---|---|
| Where ideas live | On your device, local-first | Synced across your desktops |
| AI setup | Bring your own key, Claude Code/Codex subscription, or local models via Ollama | Built-in frontier AI, no keys |
| Cost model | Free, open source (MIT) | 7-day free trial, then paid with usage limits and optional top-ups |
| Best for | Privacy-conscious tinkerers who already have model access | People who want sync and AI handled without setup |
The workflow only pays off if you actually capture consistently — a graph of one session won't remind you of much. The free desktop route suits someone who values local control and doesn't mind configuring a model; Pro removes that friction at the cost of cloud sync and a subscription.
Next step: pick one recurring question you've rethought more than twice, run a single session on it, and note how the graph links it the next time you return. That first round-trip is the cheapest test of whether the loop fits how you think.
Cloudflare
Google Workspace
User reviews (0)