How Kompas Turns Captured Ideas into a Connected Graph and Then into Writing

Kompas works as a three-stage pipeline: you capture fragments into sessions, each session is automatically mapped into a graph that links notes, references, and decisions, and the Writer draws on that graph so your output is built from your own prior material rather than starting from scratch. This applies whether you use the free desktop version (local-first, your own AI key or local models) or Pro Cloud (cross-device sync, built-in AI). The mechanism is the same; what changes is where your data lives and which AI does the work.

Stage 1: Capture — fragments land somewhere findable

The starting point is deliberately loose. Kompas is built for "your reading, your conversations, your half-formed ideas," gathered in one place. In practice, a capture can be:

• A passage from something you read
• A fragment from an AI conversation
• A note or a decision you jot down mid-thought

Captures are organized into sessions — the interface shows a session list (e.g., "Sessions / On patience") alongside Capture, Projects, Library, Writer, and Graph views. The point of capturing inside a session rather than a blank note is that the session becomes the unit that later gets connected and resurfaced.

What to expect: a fragment you save is retrievable later, not lost to a scrolling feed. The site's framing is explicit — "Every session lands somewhere you can find again, connected to what you wrote before."

Stage 2: The graph — sessions become linked structure

Once captured, a session is mapped into a live session map. The site's example session ("On patience") shows the map connecting nodes such as Patience, Camus, 009, recall, ritual, Sisyphus, and waiting. Alongside the map sits a References panel listing the specific items behind the session — for example:

Two things make this more than a tag cloud:

1. References resurface earlier thoughts. In the example, the system brings back session 009 with the line "You have circled this before — in session 009, you wrote that retrieval is the act of remembering you already chose." That is the graph doing retrieval work: connecting the current session to a prior one by content, not by manual linking.
2. The map grows across sessions. The site describes the arc as "Your sessions become a graph. Your graph becomes your writing" — the graph is the persistent layer between capture and output.

Verification step: after a few sessions, open the Graph view and check that your new session's nodes connect to earlier material. If a fragment isn't appearing in References, it likely wasn't captured as a distinct item (note, decision, or reference) within the session — recapture it as its own entry.

Stage 3: Writer — the graph feeds your draft

The Writer is where the graph pays off. Because your sessions, notes, and decisions are already linked, the Writer can draw on your own prior material when you draft, so the output "draws on your own prior material" instead of generic model output. The site's stated goal is work "that is unmistakably yours," and its framing of originality is "recollection with intent" — the tool helps you recall and reuse what you already thought.

How to use it in practice:

1. Work through a few sessions on a topic so the graph has material to draw on.
2. Open the Writer and start drafting on that topic.
3. Expect the draft to reflect your earlier notes and decisions; where it doesn't, add the missing fragment as a capture and return to the Writer.

Common snag: a thin graph produces thin writing. The Writer can only draw on what you've captured, so the first sessions on a new topic will feel sparse. This is a sequencing issue, not a tool failure — capture first, write second.

Free desktop vs. Pro Cloud — same pipeline, different setup

Both versions run the capture → graph → Writer flow. The difference is operational:

The site's own guidance: "Start free. Move to Pro when you want sync and AI handled for you." Note that the free version requires you to supply an AI provider or run local models — it is free software, not free AI usage. Pro's included AI usage has stated limits, with top-ups available; the page does not publish specific prices, so check the app for current terms.

What the design assumes about you

Kompas is aimed at "people who want their ideas to grow, not just be saved." The pipeline only compounds if you actually capture across reading, conversations, and stray thoughts, and return to sessions rather than treating them as an archive. The site also states your material is "never sold, never used to train models" — relevant if you're deciding between the local-first free version and cloud sync.