MinerU
Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
Export signed-in ChatGPT, Grok and Gemini chats to local Markdown, PDF and JSON.
作者提供的原文
教 AI 做这类工作的 Skill,可与本项一起使用
Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
Agent skills for Obsidian. Teach your agent to use Obsidian CLI and open formats including Markdown, Bases, JSON Canvas.
列表按使用分(满分 100)排序,由以下部分组成
| 热度 | 0/80 | GitHub 星标或每周下载量:较大的计 60,较小的计 20 |
|---|---|---|
| 新鲜度 | 12/12 | 30 天内有发布或代码更新得满分 |
| 仍在响应 | 0/8 | 上次检查时云端端点有响应 |
| 使用分 | 12/100 |
质量: 基础检查通过 3/3 项(仅展示,不用于排序) · 上升势头: 暂无数据 (只用于“上升中”标签)
选择你使用的 AI 应用,然后复制设置。
通过 3/3 项检查 · 在你的电脑上运行
第一步:你用哪个 AI?
选择你用的应用,步骤会显示在下方。
可连接 Claude,但需要先准备
不需要密钥,不用安装程序 · 仅限电脑上的 Claude 应用(Windows 或 Mac),网页和手机不可用
用 Convo Vault 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Claude 说正在使用 Convo Vault,或回答中出现 Convo Vault 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同
ChatGPT 暂时无法连接 Convo Vault
Convo Vault 需要安装在你的电脑上,而 ChatGPT 只能连接在网上运行的服务器
Claude Code 暂时无法连接 Convo Vault
开发者只提供了电脑版 Claude 应用的安装文件
A · 使用能连接 Convo Vault 的应用
Cursor 暂时无法连接 Convo Vault
开发者只提供了电脑版 Claude 应用的安装文件
A · 使用能连接 Convo Vault 的应用
VS Code 暂时无法连接 Convo Vault
开发者只提供了电脑版 Claude 应用的安装文件
A · 使用能连接 Convo Vault 的应用
依据这个服务器的连接方式和各应用自己的文档,检查于 2026年11月10日。
在你的电脑上运行,适用于
ClaudeGooseLM StudioMsty StudioCherry StudioVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatRaycast
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月20日
注册名称已通过 GitHub 账户 amethyst-cosmos 验证,该账户拥有此仓库
代码最后更新于昨天
使用 MIT
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/servers/amethyst-cosmos-convo-vault)提问、分享技巧或报告无法使用。按“有帮助”数量排序。
还没有评价
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 3 项适用的基础检查中通过了 3 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于昨天,最新版本(0.17.2)于昨天进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 MCPB。
同一分类中有:Puter MCP, XcodeBuildMCP, Vexa, Atlassian Rovo MCP Server, Anki MCP Server
同一分类、其他作者、命令已通过检查。官方的排在前面,再按星标排序。
如何看卡片:↓ 每周下载量i★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Puter MCP enables AI tools to interact with Puter: manage files, websites, workers, and more
XcodeBuildMCP provides tools for Xcode project management, simulator management, and app utilities.
Meeting bot and transcripts for Google Meet, Teams and Zoom. Live or after, speakers labelled.
Connect to Atlassian Jira, Confluence, Loom, and more to search, create, and manage your work.
MCP server for Anki flashcards: adaptive review, notes, media, and deck management via AnkiConnect.
| 本项 Convo Vault | Puter MCP | XcodeBuildMCP | Vexa | Atlassian Rovo MCP Server | Anki MCP Server | |
|---|---|---|---|---|---|---|
| 运行位置 | 需要安装在电脑上 | 在云端运行,无需安装 | 需要安装在电脑上 | 在云端运行,无需安装 | 在云端运行,无需安装 | 需要安装在电脑上 |
| 需要密钥 | 否 | 否 | 否 | 是 | 否 | 否 |
| 可在 Claude 中按 + 添加 | 否 | 否 | 否 | 否 | 是 | 否 |
| 基础检查 | 3/3 | 4/4 | 3/3 | 2/4 | 4/4 | 3/3 |
| 许可证 | MIT | AGPL-3.0 | MIT | Apache-2.0 | Apache-2.0 | MIT |
| 价格 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 |
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
English · 简体中文
Recommended: ego lite · Quick start · What you get · CLI · Privacy · Changelog
Current version: 0.17.2
Convo Vault is a chat exporter. It turns a conversation from ChatGPT, Grok or Gemini into a folder you own. Tell Claude or Codex to export a conversation, or send it a screenshot, and you get typeset Markdown and PDF plus structured data you can feed to agents, a knowledge base or a RAG pipeline: JSONL messages, QA pairs, a topic and entity index, and the model's full thinking and tool trace.
Everything runs on your machine. Convo Vault reads the chat tab you are already signed in to (in ego lite directly, in Chrome, Edge or Brave through a small extension), and a small local service renders the files. No cloud converter, and no cookies or tokens handed to an agent.
[!TIP]
For the smoothest experience, use Convo Vault with [ego lite](https://lite.ego.app/). There is no extension to install, no Developer mode and nothing to click. Sign in to your chat site in ego lite once; after that, telling your agent "export this conversation" is all it takes.
ego lite is a Chromium browser that lets local programs work in its pages. Convo Vault uses that to run its exporter right inside the chat tab you are signed in to, so the one manual step other browsers need, loading an unpublished extension by hand, is gone. The exported files are the same either way.
| ego lite (recommended) | Chrome, Edge, Brave | |
|---|---|---|
| Browser extension | Not needed | Load it once by hand (Developer mode, Load unpacked) |
| First time | The agent installs Convo Vault; you sign in to the chat site | The agent installs Convo Vault; you load the extension and sign in |
| After an update | Runs the new version at once | The extension reloads itself |
| What you get | The same folder | The same folder |
When ego lite is installed, the CLI and the skill use it automatically; --browser ego picks it explicitly. Chrome, Edge and Brave keep working with the extension.
| Chat | Give it a link like | How it is read | What the export is checked against |
|---|---|---|---|
| ChatGPT | chatgpt.com/c/ | its conversation API | its own message list, by message id |
| Grok | grok.com/c/ | its conversation API | its response tree |
| Gemini | gemini.google.com/app/ | the page itself, scrolled to the top first | only what the page shows (a weaker check, stated in a warning) |
Every chat gets the same folder: readable Markdown, a typeset PDF, messages.jsonl, question-and-answer pairs, a summary, the thinking trace where the site has one, and the images. What each site cannot do yet (attachments, list, --project) is listed under Grok and Gemini below; ChatGPT does everything.
Agent-native A skill for Claude Code and Codex, an MCP server for chat apps, and a CLI. Paste a link or a screenshot; the agent installs, exports and reports. No panel to click, no mode to pick.
Nothing silently lost Every export is checked message by message against the chat's own message list (ChatGPT and Grok; Gemini can only be checked against what its page shows, and says so). A gap blocks the export and tells you what is missing. Images, citations, thinking steps and generative UI are kept.
Local by design The renderer listens on 127.0.0.1 only, jobs can only be claimed by the extension or a local program (never a web page), and your login stays in the browser. Files never leave the machine.
In Claude Code, the Code tab of the Claude desktop app, or Codex, send:
Install https://github.com/amethyst-cosmos/convo-vault for me. Follow INSTALL.md in the repository and put it in ~/convo-vault.
The agent clones the repository, installs dependencies, builds the extension, installs the skill and runs a check.
ego lite (recommended): just sign in. Sign in to your chat site (ChatGPT, Grok or Gemini) in ego lite. There is nothing to install in the browser; Convo Vault picks ego lite by itself (or pass --browser ego).
Chrome, Edge or Brave: load the extension once. These browsers do not let programs install unpublished extensions, so this one step is manual:
chrome://extensions or edge://extensions) and turn on Developer mode.~/convo-vault/dist/convo-vault-extension.Start a new agent session and say:
Export this conversation: https://chatgpt.com/c/xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
A Grok (https://grok.com/c/…) or Gemini (https://gemini.google.com/app/…) link works the same way.
Or send a screenshot of the conversation and say "download this whole conversation". The agent reads the link from the address bar. The folder appears in ~/Downloads/convo-vault/.
Prefer to install it yourself?
Download convo-vault- .zip from Releases and unzip it, or clone the repository, then:
node cli/convo-vault.mjs setup # install dependencies, build the extension, install the skill
node cli/convo-vault.mjs verify # after loading the extension: confirm the browser answers
Requires Node 20 or newer.
| Environment | Works | How |
|---|---|---|
| Claude Code, the Code tab of the Claude desktop app, Codex | ✅ | Skill, installed by setup |
| Claude desktop chat, Cursor and other clients with local MCP | ✅ | node cli/convo-vault.mjs mcp install, then send a link or screenshot in chat |
| claude.ai, chatgpt.com, grok.com and gemini.google.com in the browser (including ChatGPT agent mode) | ❌ | They run in a cloud sandbox and cannot reach your browser or local services |
The skill is the self-contained folder skills/convo-vault-skill. Copy it anywhere, or run node cli/convo-vault.mjs skill install --dir . Its launcher finds the project, or installs it from GitHub the first time.
The same skill is listed, free, on Agensi, a marketplace that serves skills to your agent over MCP. It is a text-only version (skills/convo-vault-agensi): on first use the agent asks your permission, installs this project from GitHub into ~/convo-vault, and you load the extension once. It is the same open-source code either way; the Agensi listing is only another way to find it.
Grok conversations work the same way: give the agent a https://grok.com/c/ link, or run node cli/convo-vault.mjs export https://grok.com/c/ . The extension reads Grok's own conversation API with the session your browser already has (no token, no scrolling) and checks that every message came back before it exports. The PDF and data files look the same as a ChatGPT export, labelled Grok; Grok's search steps become the thinking trace and its citations become footnotes.
Images (the ones Grok generated and the ones you uploaded) are downloaded into the bundle; other attachments are listed by name. Not yet for Grok: downloading attachments, list and --project. A conversation with regenerated answers exports its newest branch and says how many alternatives it skipped. Links to several Grok conversations run in one grok.com tab, and a mix of ChatGPT and Grok links gets one tab per site.
Gemini conversations work too: give the agent a https://gemini.google.com/app/ link. Gemini has no conversation API the extension can call, so the page itself is read: the extension scrolls the conversation to the top until no older turns appear, turns each reply's HTML back into Markdown (lists, tables, code with its language, math, links) and exports the same bundle, labelled Gemini. Several Gemini links each get their own tab, one after another.
Honest limits: Gemini does not say how many messages a conversation has, so the check is only what the page shows (a reply for every prompt, nothing still being written, no turns dropped while loading) and the export says so in a warning. Images, including the ones Gemini generated, are read from the page and embedded; if the page does not let a script read one, it appears as a link and the export says how many. The collapsed thinking panel, source chips and attachments are not exported (attachment names appear where the page shows them). This is the most fragile of the three sites: a Gemini redesign can break it.
Each export is a folder (and a .zip), including a PDF that reads like the chat page.
| File | Contents |
|---|---|
*.md | Readable Markdown with clickable citation footnotes |
*.pdf | Typeset PDF; formulas, Mermaid, charts and generative UI drawn statically |
*.summary.md | Summary index |
*.messages.jsonl | One message per line, ready for search and RAG |
*.qa-pairs.json | Question and answer pairs |
*.topics.json, *.entities.json | Topic index; links, file names, dates |
*.agent-trace.json | Full thinking and tool trace per answer, in order, with complete tool outputs and a source index |
*.agent-trace.md | Readable version of the trace |
*.assets.manifest.json | Where every image and attachment came from, how it is stored, why it degraded; size and SHA-256 of each downloaded file |
assets/files/ | Only with --files or --files-kind: the Word, PDF, zip... files uploaded to or generated in the conversation |
*.payload.json, *.conversation.json, *.data.json | Complete export data and metadata, enough to re-render later |
PDF output: formulas, Mermaid diagrams and charts, rendered from the test fixtures
flowchart LR
A["You / agent"] -->|"export link"| B["CLI"] B -->|"create job"| C