OakReader：与 AI 智能体共享的 AI 知识库

An AI knowledge base, shared by you and your agent.

Everything you read — PDFs, papers, web pages, books — in one place, then an AI research agent that searches and answers across all of it, with citations you can open.

Website · Download · 简体中文

What is OakReader

Most readers just display pages. OakReader keeps everything you read in one place — then puts an AI research agent on top that reads the whole thing for you: search across your entire library, ask questions, and get answers grounded in your sources with citations you can click open. The same knowledge base is something your AI agent can use too.

It's a native macOS app — SwiftUI + AppKit, PDFKit — built to be fast and to stay out of your way.

Features

Everything you read, one place Every PDF, paper, and web page in collections and tags — full-text search that even understands Chinese, Japanese, and Korean.	Chat with what you're reading Attach a page or selection and ask — OakReader answers right beside the document, in context, with formatted notes, code, and rendered math.
Translate inline, never break flow Select any passage to see it in your language right beside the original — sentence by sentence, no copy-paste, no jumping to a separate translation app.	Go deep on any word Click a word and get a detailed card — meaning, nuance, collocations, and a one-line takeaway.
Highlight, note, keep Mark up any page; notes and screenshot clips collect into one panel — searchable across everything you've read.	The web, built in Search, open links, and read live pages signed in — then save what matters into your library.

Use any model, switch anytime — from the frontier labs to a local model running on your Mac. Your choice, every time.

Tech stack

Component	Framework
UI	SwiftUI + AppKit
PDF engine	PDFKit
Catalog	GRDB (SQLite)
AI	Multi-provider (Claude, OpenAI, Gemini, local Ollama/LM Studio, and more) via the local `OakReaderAI` package
Search	Full-text search with CJK support
Web capture	Browser extension (WXT) → local snapshot server
OCR	Apple Vision

Requirements

macOS 15.4 or later
Xcode 16 / Swift 6 (to build from source)

Build

git clone https://github.com/oakreader/oakreader.git
cd oakreader
brew install xcodegen
xcodegen generate

Code signing (one-time, for your own build)

The committed project signs with the maintainer's Apple team, which you can't use. Point it at your own identity with a local, git-ignored override — you don't edit any tracked file:

scripts/setup-signing.sh

It asks for your Apple Developer Team ID (find it at developer.apple.com → Account → Membership; a free Apple ID works) and writes xcconfig/DeveloperSettings.xcconfig. No Apple account? Press Enter to build ad-hoc instead — fine for local development. (You can also copy xcconfig/DeveloperSettings.example.xcconfig by hand.)

Debug builds drop the restricted keychain-access-groups entitlement so any team — or ad-hoc — can build. Your API keys live in your own login keychain (enter them in Settings → AI), or you can inject them via the provider env vars, e.g. OPENAI_API_KEY.

Run

open OakReader.xcodeproj

Select the OakReader scheme, then build and run (Cmd+R). project.yml is the source of truth — the .xcodeproj is generated by XcodeGen.

Download

A signed and notarized build is available on the website and on the Releases page.

Contributing

OakReader is open source and contributions are welcome — bug reports, fixes, and focused improvements. Open an issue to discuss anything substantial before sending a large PR.