How Kanto Notes works.

1. Sign up with email + password. Free forever, no card.
2. On the New Note screen, type or hit Record and speak. Voice is auto-transcribed via Whisper.
3. Click Save. The note is saved instantly.
4. AI enrichment runs in the background (~4s). Refresh All Notes to see the auto-title, tags, and category appear.
5. Use Search to find by meaning, or Ask Notes to chat over your notes.

The home screen is intentionally minimal. One title field, one big textarea, one microphone button. Save is a single click. Latency from click to persisted row is under 200ms.

{
  "id": "a4c8...",
  "title": "Untitled",
  "content": "Reminder for the Friday demo: swap the pricing slide
with the new one Sarah sent, and mention the AWS migration savings.
Also add the customer quote from Acme.",
  "created_at": "2026-07-15T10:14:22Z"
}

The note is queryable immediately — you don't wait for AI.

Right after save, Kanto fires a server function that runs two things in parallel: (1) a chat completion that returns a strict JSON structure with title, category, tags, and entities; (2) a text-embedding-3-small call that produces a 1536-dim vector for later semantic search.

system: You organize personal notes.
Return STRICT JSON only.
{
  "title": string (<= 8 words),
  "category": "idea"|"todo"|"meeting"|"journal"|"reference"|"misc",
  "tags": string[] (1-5, lowercase),
  "entities": [{ "type": "person"|"org"|"project"|"date"|"topic",
                 "value": string }]
}

user: Reminder for the Friday demo: swap the pricing slide with
the new one Sarah sent, and mention the AWS migration savings...

{
  "title": "Friday demo prep: pricing slide + AWS savings",
  "category": "todo",
  "tags": ["demo", "friday", "pricing", "aws", "acme"],
  "entities": [
    { "type": "person", "value": "Sarah" },
    { "type": "org", "value": "Acme" },
    { "type": "org", "value": "AWS" },
    { "type": "date", "value": "Friday" },
    { "type": "topic", "value": "pricing slide" }
  ]
}

You can override anything from the note's edit view. Overrides are preserved on subsequent enrichments.

Search takes your query, embeds it with the same model, and calls the match_notes Postgres function which orders your notes by cosine distance (filtered by your user id and non-deleted rows).

{
  "id": "e12c...",
  "title": "Async standups experiment",
  "similarity": 0.82,
  "content": "Trialled a Loom-based standup for 2 weeks with the
    remote team — every morning a 90-second video, timestamps in
    Notion. Engagement was higher than the sync one..."
}

Ask Notes is retrieval-augmented chat: your question is embedded, top-K similar notes are pulled, and they're passed as grounded context to gpt-5.4-nano with instructions to cite the specific notes it used.

If the answer isn't in your notes, Kanto will say so instead of hallucinating.

The list view is optimized for large personal corpora:

• Filter by keyword, category, or tag.
• Sort by recently updated, created, or title.
• Soft-delete puts the note in a restorable Trash for 30 days, then hard-delete.
• Bulk actions on multi-select.

In Settings, you can paste your own sk-... key. From that moment on, every AI operation for your account — enrichment, embeddings, ask, transcription — is billed to your OpenAI account instead of ours.

Provider: openai
Chat model: gpt-5.4-nano
OpenAI API key: sk-proj-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Every feature works on every plan — only the monthly caps change. Limits reset on the 1st of each month (UTC).

Prices: Free $0 · Standard $2/mo · Pro $6/mo · Ultimate $14/mo. Yearly saves ~30%. See pricing for full details.