Scroll Agent

io.github.KandechoPublished by GitHub account Kandechov0.1.0Updated Nov 4, 2026

Independent Bilibili account, personal feed, inbox and authorized delivery; automation in trial.

VerifiedSTDIODesktop onlyKnowledge & MemoryMedia & Design

Overview

AI-generated overview

Lets an AI assistant read Bilibili, YouTube and local videos — subtitles, audio and sampled frames — to summarize, verify details and discuss with timestamps.

What you can do
  • Read videos: pull subtitles, speech and sampled frames from Bilibili, YouTube or local files so the assistant can summarize points and arguments.
  • Check details: revisit a specific timestamp and compare what is shown with what is said.
  • Continue and export: resume a partially read video and export cited frames and quotes.
  • Discover and preview: find Bilibili videos by topic and preview storyboards.
  • Read engagement: fetch comments, replies and danmaku on demand.
  • Manage a queue and history, plus an optional Scroll workflow for a separate Bilibili account's feed, inbox and authorized delivery.
When to use it
Useful when you want an assistant to work through a video link directly — summarizing, fact-checking a moment on screen, or picking up a long video in parts — instead of writing notes first. The Scroll feed and inbox features are optional and not needed for ordinary reading.
Requirements
Local Windows x64 install with PowerShell 7 and Python 3.12; needs uv and FFmpeg/ffprobe, plus Node.js for YouTube. Bilibili access uses the component's own QR-code login; YouTube can use a manually exported short-lived cookie. Local transcription must be configured separately for videos without usable subtitles.
Before you install
Platform logins and cookies are involved: Bilibili via QR login, YouTube via an imported cookie, and Scroll can sign in a separate account that sends and receives authorized private messages. Cloud reading via Gemini is off by default but can be enabled; material is otherwise processed by your chosen host and model. Subtitles and machine recognition can be wrong, and sampled frames may miss brief actions.

Installation

In SourceWeft

  1. Open Scroll Agent in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Video Reader

把 Bilibili、YouTube 和本地视频交给你正在用的 AI 助手阅读：整理内容、核对画面，或带着时间点继续追问。

Video Reader 提供字幕、语音和画面材料，通过 MCP 接入 Codex、Claude Code 和 DSH；理解和回答由助手完成。你可以直接贴链接提问，不必先生成笔记。

开始使用 · 下载整包 · 文档

可以用来做什么

  • 读懂一条视频：“整理这条视频的主要观点和论据，标出值得回看的位置。”
  • 核对具体细节：“06:12 的图和讲解一致吗？把相关画面拿出来看。”
  • 接着讨论或保存：“继续上次没看完的部分，把引用的画面和原话一起导出。”

还可以按主题找 B 站视频、预览分镜，以及按需读取评论、回复和弹幕。长视频可以分次阅读，已取得的材料可继续使用。

整包还包含通用队列与历史、Scroll 个人工作流。Scroll 可给组件单独登录 B 站小号，使用小号首页推荐，与主号收发授权私信、接收转发视频并入队。这些能力已有真实使用；定时工作流等仍在试用，常驻收件和持续后台阅读尚未接入。详见能力状态。

开始使用

当前发布版面向 Windows x64、PowerShell 7 和 Python 3.12。安装需要 uv、FFmpeg / ffprobe；读取 YouTube 时还需要 Node.js。uv 可按需准备 Python。

下载并解压源码包，在解压后的目录打开 PowerShell 7，运行：

powershell
& './bootstrap.ps1'& './run.ps1' doctor

然后按 首次使用指南 接入你的助手，贴一个视频链接并说明你想了解什么。安装不会自动下载语音模型；没有可用字幕的视频，需要另行 配置本地转写。

先了解这些边界

  • 处理默认在本地进行。 Gemini 云端读取默认关闭，不会在失败后自动切换到云端。交给助手的材料由你选择的宿主和模型处理。
  • 字幕和机器识别可能出错。 重要细节可以回到原话、原图核对；取样画面不能保证覆盖每个短暂动作。
  • 平台访问需要相应权限。 B 站使用组件自己的扫码登录；YouTube 可导入你手动导出的短期 Cookie。登录和平台限制仍可能影响读取。

本版的已知限制和验证范围见 发布说明。

按需要查文档

你想做什么从这里开始
第一次安装、接入助手、读一条视频首次使用
在 DSH 中安装整包薄适配DSH 接入
从 DSH 插件索引发现本项目发现与收录 · 开放插件市场条目
使用 MCPB 接入包、在官方目录查询MCP Registry
区分可用能力、试用部分与未完成项能力状态
设置模型、登录或调整文件位置安装与配置
查命令、工具参数、续读和导出CLI 与 MCP 参考
管理待读视频、附言和结果队列与历史
使用个人推荐、私信收件和授权投递Scroll
报告问题或参与开发贡献与复现
了解开发顺序与阶段验收开发路线

Scroll 可按需使用，普通读片无需配置个人投递。文档首页列出完整入口；与 dsh-bili-scroll 等实现的不同侧重见相关项目。

许可

项目原创代码采用 MIT。第三方代码、依赖和模型保留各自许可，见 第三方声明。源码包不含用户视频、私信、凭据或模型。

Source: README.md at commit 4a9f981

Tools

0
SourceWeft started this package once in an isolated sandbox, without credentials, to list its tools. That failed: the package type cannot run in the sandbox.

Version history

1
  1. v0.1.0LatestNov 4, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts