Grape
Handwriting notes for developers

Write it by hand. Ask it like text.

Write debugging sessions and design discussions by hand on pages of ruled paper and Grape reads the words off the page on the device itself, so a page of a pile of runbooks and notes in your own hand turns up in search like any typed note and the agent can quote it.
Lined, grid, dotted, graph or plain pages, A4 or Letter, up to a hundred of them
Handwriting read into text on the device, so pages are searchable and the agent can read them
Import a PDF to write over it, export the notebook as one
The same pages open on the desk, where you can keep writing on them

Paper that behaves like paper

Pen, highlighter, eraser, shapes and text boxes on pages that stack the way a notebook does.

Your handwriting, searchable

The words are read off the page on the device itself, and nothing leaves it to do so.

The agent reads your pages

Ask what the lecture said on Tuesday and the answer comes from the ink.

Write on the slides

Import a PDF and annotate it page by page, then export the lot back out as one.

Page forty is a tap away

A navigator shows the whole notebook as thumbnails, with a page added or removed from there.

Open the notebook at the desk

The pages sync like any note and open in the desktop app, where you can add to them.
// results

Three bugs in the workflow, closed

No feature tour. What changes when your runbooks can be queried instead of grepped.
1.
You fixed this bug in March and are debugging it again from scratch.

Stop solving the same bug twice

Describe the symptom in search and get the note where you already worked it out.
2.
The decision lives in a Slack thread that scrolled away.

Keep decisions where you can defend them

Ask what you decided about a service and get the note, the date and the reasoning.
3.
The AI tool wants your notes on its server, and you want them on disk.

Point the AI at your own stack

Bring your own key, or run a local model through Ollama, and keep the workspace on disk.
// usage

What that looks like for developers

Three points in a sprint where the tool earns its slot.

Debugging

Write the symptom, the theory and the fix in one note, with the log you kept attached.

Architecture review

Record the conversation and keep the decision log next to the diagram.

New project

Keep the setup steps in a runbook that actually gets found again.
// faq

Questions Before You Install It

The rest of them live in the help center.
Can I run the AI locally?
Yes. Grape connects to Ollama and LM Studio as well as OpenAI, Anthropic, Google, Groq, DeepSeek, Grok and OpenRouter, and you can add any OpenAI compatible endpoint yourself.
How does it handle technical notes?
Inline code, tables, checklists, Markdown paste that keeps formatting, and export back out to Markdown when you want the note somewhere else.
What is Grape, in one paragraph?
Grape is a note app for Mac, Windows, Linux, iPhone and Android with AI built into the middle of it rather than bolted on the side. You write, record and collect things the normal way, and then you can search by meaning, ask questions about your own notes, and turn any note into a summary, a set of flashcards, a quiz or a mind map.
What makes it different from other note apps?
Most note apps treat AI as an extra panel. In Grape the agent works on your workspace: it can read, create, edit, move, tag and delete notes, and every answer points back to the note it came from. Semantic search, voice transcription, flashcards, quizzes, mind maps and the drawing canvas all live inside the editor rather than in separate tools.
Where do my notes actually live?
On your device, in a local database, and you can choose where that file sits. Nothing needs to leave your machine for the app to work. If you turn on sync with a Pro plan, your notes are also stored in the cloud so your devices can stay in step.
Do I need an account to use it?
No. Download the app and start writing. You only need an account for the paid plans, which is where built in AI and cross device sync come in.

Find it next time it breaks.

Runbooks, architecture decisions and the fix you worked out at midnight, in a local library you can search by meaning and keep offline.