v0.2.2 · a fork of Logseq with an LLM retrieval layer · github↗

Don’t browse your notes.
Peck them. Get answers.

Kip is the full Logseq editor — Markdown notes, journals, whiteboards, block references, unchanged — plus a layer that turns documents you drop in into a cross-linked wiki you can ask questions of. It opens straight into a chat prompt; the editor is a mode you toggle into.

Star on GitHub ↗

windows 10/11 x64 · linux x64 · free · agpl-3.0

hatcheggs/ → the nest
kip — peck
what did we decide about the timeline?
The launch moved to [[Q3 Planning]] after the vendor delay noted in [[Acme Meeting 06-12]]. Design freeze stays on the 18th — see [[Timeline]].
answered from 3 pages in your nest⌘↵
peckask · tell · delegate
groomread-only health checks
01 — the lifecycle of an idea

Drop an egg. Hatch a nest. Peck for answers.

Every document you feed Kip becomes part of a cross-linked wiki — “the nest” — built from entity, concept and source pages that cite each other.

EGG

Drop it in

Put any .md or .txt file into eggs/ inside your graph folder. That’s the whole import flow.

HATCH

It becomes pages

Hatch sources turns each document into a set of linked entity, concept and source pages — the nest.

PECK

Ask the nest

Type a question; answers cite the [[pages]] they came from. Or tell Kip a fact, or about an upcoming meeting.

GROOM

Keep it tidy

Read-only health checks run over the generated wiki, so the nest stays trustworthy as it grows.

02 — what’s inside the shell

Chat-first, editor always one key away.

the editor

The whole Logseq, unchanged.

Kip doesn’t reinvent your notes — it inherits them. Everything you know from Logseq works exactly as you expect.

  • Markdown notes & journals, outliner and all
  • Whiteboards for when text runs out
  • Block references and backlinks, untouched
  • Kip opens into a chat prompt — press Ctrl/⌘ + 1 to toggle the editor
your graph/
# everything Kip makes lives in your folder
graph/
├─ eggs/        ← drop documents here
├─ nest/        ← hatched wiki pages
│   ├─ entity___Acme.md
│   ├─ concept___Timeline.md
│   └─ source___q3-plan.md
├─ clucks/      ← Kip's working notes
├─ journals/
└─ pages/
the skill loop

Pecks that do things.

Mid-answer, a bounded skill loop can run real tasks against the real world — then fold the result back into the reply.

  • Search the web while answering you
  • Read a spreadsheet from your disk
  • Build a Word doc or a deck from your notes
  • Bounded by design — and your own skills run with your privileges, so treat them like shell scripts
peck — with skills
 make a deck of the Acme decisions

→ reading [[Acme Meeting 06-12]]
→ reading [[Q3 Planning]]
→ skill: build-slides (node subprocess)

 acme-decisions.pptx — 9 slides,
  every claim cited to a page
reminders

It remembers before you need it.

Tell Kip what’s coming in plain words — it sets an OS notification and, when the moment approaches, hands you a prep brief pulled from your own notes.

  • “I have a meeting with Acme on Friday at 15h” — done, that’s the whole interface
  • OS notification beforehand, no app-checking required
  • Prep brief assembled from the relevant pages in your nest
notification — friday 14:30
Kip · reminder
Acme meeting in 30 min

Prep brief from your nest:
• Timeline moved to Q3 ([[Q3 Planning]])
• Budget flagged twice ([[Acme Meeting 06-12]])
• Open question: renewal pricing
privacy

Your nest, your machine.

Your notes and questions go to whichever LLM provider you configure — Anthropic, OpenAI and DeepSeek are hosted services that receive that content. Don’t want that? You don’t have to have it.

  • Local provider (Ollama) keeps everything on-device
  • Your graph is a plain folder — plain Markdown, no lock-in
  • Honest bit: .henhouse/llm.json stores API keys in plaintext — keep that folder out of synced locations and version control
settings → llm
provider:
 ( ) anthropic   — api key
 ( ) openai      — api key
 ( ) deepseek    — api key
 () local       — ollama, on-device

✓ connection tested — all traffic stays
  on this machine
03 — install

No installer. No account. Just unzip and peck.

Kip ships as a plain folder archive from GitHub Releases. To update later, download the new release and replace the folder — there’s no auto-updater yet.

Windows

10 / 11 · x64

Download the zip, extract it anywhere, run Kip.exe.

↓ Kip-windows-x64.zip
the binary is unsigned — SmartScreen will say “Windows protected your PC”. Click More info → Run anyway.

Linux

x64 · tar.gz

Extract and run — plus a launcher with Wayland flags baked in.

↓ Kip-linux-x64.tar.gz
tar -xzf Kip-*.tar.gz && ./kip
Wayland (Hyprland etc.): ./kip --ozone-platform-hint=auto
04 — first five minutes

From zero to first peck.

  1. Open a folder as your graph

    Everything Kip creates — eggs/, nest/, clucks/ — lives inside it.

  2. Set an LLM provider

    Settings → LLM: Anthropic, OpenAI or DeepSeek (API key), or Local via Ollama. Hit Test connection.

  3. Drop a document

    Put a .md or .txt file into <graph>/eggs/.

  4. Hatch it

    Header “…” menu → Hatch sources → Start. It becomes pages under The Nest.

  5. Peck

    Type a question in the prompt. Answers link back to the pages they came from.

  6. Toggle the editor

    Ctrl/⌘ + 1 whenever you want to read or write notes directly.

05 — said plainly

v0.2: early, rough, and looking for feedback.

Kip is a personal project at the beginning of its life. Here’s exactly what that means, so you can decide with open eyes.

Windows + Linux x64 only. No macOS build, no mobile.

Unsigned binaries, no auto-update. Folder-zip, not an installer.

Needs an LLM provider configured — without one, Hatch and Peck don’t work.

Skills run unsandboxed with your privileges — one you add yourself is like running a shell script.

Expect bugs and breaking changes. It’s v0.2, and it means it.

Which is where you come in. Bugs and ideas go to Issues; questions and general feedback to Discussions.

Come peck around.

Free, open source (AGPL-3.0), and one folder away from knowing what your notes would say back.

View the repo ↗