JinaClip is a local-first browser extension that converts any webpage into LLM-friendly Markdown, JSON, or HTML inside the browser runtime, aimed at people who need webpage content for AI workflows without a cloud service.

What it is

The README presents the project as Jina Reader — Local Edition (v7). It is a JavaScript project under the MIT license, packaged as a Manifest V3 browser extension, and organized as a modular service-oriented architecture with background orchestration, LocalReaderService v7, an intelligence engine, logging utilities, and configuration modules. The repository names manifest.json, scripts/build.js, src/background/index.js, src/services/localReader.js, src/utils/logger.js, src/config/constants.js, and src/config/defaultSettings.js.

JinaClip solves the problem of extracting clean, LLM-friendly content from modern webpages that use Shadow DOM, dynamic iframes, and SPA layouts without sending the page to a remote processor. It lives in the browser-extension ecosystem, specifically the Chrome extension environment indicated by Manifest V3 and the chrome-extension and mv3 topics, and it replaces the cloud path through r.jina.ai by implementing extraction and transformation in the browser runtime.

Key capabilities

• Converts captured page content into Markdown, JSON, or HTML and writes the result to the system clipboard through the Clipboard API, with Notification API feedback.
• Uses a Heuristic Scoring Engine that calculates a text-to-link density ratio and penalizes interactive noise such as dashboards, buttons, and navigation.
• Flattens Shadow DOM and dynamic iframes during extraction; the feature matrix marks Shadow DOM as supported in the Local Engine (v7).
• Captures full pageshots by analyzing scrollHeight and viewHeight, scrolling with paint-settle delays, assembling segments in an OffscreenCanvas, and returning a Markdown-wrapped Base64 string.
• Exposes copy actions through click, hotkeys, and browser menu entry points, routed by a background orchestrator and config service to LocalReaderService v7.
• Packages as a Manifest V3 extension through manifest.json and scripts/build.js, with a vanilla ES6+ codebase.
• Marks table formatting as Pro Grade and full pageshots as Stitched Canvas in the feature matrix, alongside 100% local processing, infinite rate limits, and a $0 cost model.

Who uses it and how

• AI workflow users who need webpage content in prompts can trigger a copy action and receive Markdown, JSON, or HTML on the clipboard without a cloud round trip.
• Privacy-centric users can keep page content in the browser runtime, as described by the local-first and privacy-centric topics.
• Users dealing with SPA layouts, Shadow DOM, and dynamic iframes can rely on the DOM scoring and flattening pipeline.
• Chrome extension users in a Manifest V3 environment are the deployment audience indicated by manifest.json, the chrome-extension topic, and the mv3 topic.
• Contributors work in vanilla ES6+ and use the scripts/build.js packaging pipeline; the README's installation and development section is contributor onboarding.

Getting started

The README provides a contributor-oriented installation and development path: the codebase is vanilla ES6+, packaged through scripts/build.js, and defined as a Manifest V3 extension by manifest.json. No Docker image, package-manager command, compose file, or hosted option is given in the supplied facts.

How it compares

Within the supplied facts, JinaClip is the local-first counterpart to Jina Reader's Cloud Reader, removing the need for r.jina.ai cloud calls by running extraction and transformation in the browser runtime. The feature matrix contrasts the Local Engine (v7) with Cloud Reader on privacy, rate limits, cost, Shadow DOM support, table formatting, and full pageshots.

When to use it — and when not to

JinaClip is a fit for users who want webpage-to-Markdown conversion for LLM workflows without operating a server and who are comfortable with a Manifest V3 extension built through the repository's contributor path. It is not a fit for anyone who needs a documented hosted service, a Docker deployment, or a package-manager install, because the supplied facts provide none of those. The README excerpt is truncated and does not document a store listing or release artifact, and the registry category lists System Cleanup & Optimization even though the documented function is webpage conversion.

graph TD
    User([User Action: Click/Hotkeys/Menu]) --> Orchestrator[Background Orchestrator]
    Orchestrator --> Config[Config Service]
    Config --> Service[LocalReaderService v7]
    
    subgraph "Target Tab Context"
        Service -- Script Injection --> Engine[v7 Intelligence Engine]
        Engine -- DOM Scoring --> Root[Content Root Discovery]
        Root -- Cleaning --> Clone[Sanitized Clone]
        Clone -- Conversion --> MD[Markdown/JSON/HTML]
    end
    
    MD -- Message Passing --> Service
    Service -- Clipboard API --> OS[System Clipboard]
    Service -- Notification API --> User

sequenceDiagram
    participant P as Web Page
    participant E as Intelligence Engine
    participant S as Scoring Module
    participant C as Converter
    
    E->>P: Traverse DOM Tree
    P-->>E: Candidate Elements (div, main, article)
    E->>S: Compute (Text Density - Link Density)
    S-->>E: Final Scores
    E->>P: Flatten Shadow DOM & Iframes
    E->>C: Execute Pro-Grade Markdown Conversion
    C-->>E: LLM-Friendly Payload

p:/extensions/JinaClip - Copy Page for LLM/
├── scripts/
│   └── build.js          # Professional packaging pipeline
├── src/
│   ├── background/
│   │   └── index.js      # Lifecycle & Event Orchestration
│   ├── services/
│   │   └── localReader.js # Core Extraction & Transformation Engine (v7)
│   ├── utils/
│   │   └── logger.js      # System-wide observability
│   └── config/
│       ├── constants.js   # Immutable definitions
│       └── defaultSettings.js
├── icons/                # High-fidelity visual assets
└── manifest.json         # Extension Manifest (MV3)