PluginBench
MCP Server
Active
MIT

Obsidian Vault MCP MCP Server

io.github.luffysolution-svg/obsidian-vault-mcp

Local research workflow MCP: Zotero sources → MinerU extraction → Obsidian vault with structured analysis and traceable AI skills.

What is the Obsidian Vault MCP MCP server?

The Obsidian Vault MCP is a local Model Context Protocol server that integrates Zotero, MinerU, and Obsidian to create a complete research literature workflow. It provides 31 MCP tools and 7 AI skills for importing papers, extracting full text, organizing literature notes, and performing structured analysis—all with transactional safety and version verification.

This server bridges Zotero (reference management), MinerU (PDF-to-Markdown extraction), and Obsidian (knowledge vault) into a unified research system. It automates literature import, full-text parsing, wiki synthesis, and five types of structured analysis (full_read, literature_review, passage_qa, figure_qa, concept). AI agents can execute traceable research workflows via built-in skills while maintaining data integrity through dry-run, staging, and atomic transactions.

How to install Obsidian Vault MCP

Copy-paste configuration for popular MCP clients.

transport: stdio
Config generated by PluginBench — verify against the source before use.
Environment / auth
  • OBSIDIAN_VAULT_PATH
    required

    Absolute path to the Obsidian vault, or auto when the client project is inside it.

~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "obsidian-vault-mcp": {
      "command": "uvx",
      "args": [
        "zotero-obsidian-mcp",
        "serve",
        "--transport",
        "stdio"
      ],
      "env": {
        "OBSIDIAN_VAULT_PATH": "<YOUR_OBSIDIAN_VAULT_PATH>"
      }
    }
  }
}

Tools & capabilities

Tools this server exposes to the agent.

  • literature_version — Returns current version, 31 tools, 7 skills, and five analysis types for verification.
  • zotero_search_items — Search Zotero library by query.
  • import item — Import single paper from Zotero by key with support for notes, annotations, and attachments.
  • mineru parse — Extract and normalize PDF to Markdown using MinerU with image organization.
  • Literature/index.md — Auto-maintained literature index with metadata.
  • Literature/Literature.base — Obsidian database view of all literature.
  • Analysis/Analysis.base — Unified database view with 9 perspectives on research analysis.
  • paper-qa — Single-paper quick Q&A without vault writes.
  • full-read — Complete single-paper reading with full_read analysis saved.
  • passage-qa — Locate and answer questions about specific passages, methods, data, or conclusions.
  • figure-qa — Interpret figures, tables, schemes, and equations.
  • compare-papers — Build comparative matrix across selected papers.
  • literature-review — Thematic literature review across paper pool.
  • concept-learning — Build concept models across multiple papers.

Use cases

  • Import papers from Zotero, extract full text via MinerU, and organize them in Obsidian with automatic indexing.
  • Perform structured analysis on papers: full reads, passage Q&A, figure interpretation, and concept learning with traceable AI workflows.
  • Build comparative matrices across multiple papers and synthesize literature reviews with cross-referenced wiki pages.
  • Query specific passages, methods, and data points across your literature vault using AI-assisted passage location.
  • Maintain a knowledge base where papers, analyses, and concepts are interconnected and queryable through Obsidian's database views.

Obsidian Vault MCP MCP server FAQ

What does the Obsidian Vault MCP do?

It integrates Zotero, MinerU, and Obsidian into a research workflow. You manage sources in Zotero, extract full text via MinerU, and organize everything in Obsidian with 31 tools and 7 AI skills for literature analysis, synthesis, and concept learning.

Is it free?

Yes. The server is open-source and available via PyPI (zotero-obsidian-mcp). However, MinerU (PDF extraction) may send data to external services—verify authorization and organizational policy before use.

How do I install it in Claude or Cursor?

Install via uv, pipx, or pip: `uv tool install zotero-obsidian-mcp==3.0.2`. Configure in Claude/Cursor settings with the MCP registry ID `io.github.luffysolution-svg/obsidian-vault-mcp` or use the provided stdio JSON config with your vault path.

What authentication is required?

You need Zotero Desktop running locally with the API enabled. Optionally configure MinerU token and linked attachment base directory. No cloud credentials required for core functionality.

What Python version is needed?

Python 3.10 or higher. Install with uv (recommended), pipx, or pip.

Does it support multiple clients?

Yes. The server works with Codex, Claude Code, OpenCode, Pi, Hermes, and WorkBuddy. Each client can install native plugins and skills via the `agent install` command.

README (reference)

Source of truth, from the repository.

<!-- mcp-name: io.github.luffysolution-svg/obsidian-vault-mcp -->

Obsidian Vault MCP

面向科研文献工作流的本地 MCP 服务:以 Zotero 管理来源,以 MinerU 提取全文,以 Obsidian 沉淀文献、Wiki 与结构化 Analysis,并通过 Skills 让 AI Agent 按可追溯流程工作。

English · 完整安装教程 · 开发文档 · 更新日志 · 贡献者

架构

用户自然语言任务
        ↓
7 个科研 Skills:识别意图、规划步骤、约束证据与输出
        ↓
31 个 MCP Tools:版本契约、查询、导入、解析、检索、校验与事务写入
        ↓
Zotero Desktop ── PDF ── MinerU ── Obsidian Vault
                                      ├─ Literature 主笔记
                                      ├─ PDF 与全文 Markdown
                                      ├─ Index / Literature.base
                                      ├─ Wiki
                                      └─ 五类 Analysis / Analysis.base

项目不绑定大模型供应商。MCP Tools 负责确定性的本地数据操作,Skills 负责把工具编排成可复用的科研工作流。

核心功能

  • 稳定文献身份:以 Zotero 父条目 zoteroKey 作为主键。
  • Zotero 导入与同步:支持单篇、Collection、notes、annotations、BibTeX、存储附件和链接附件。
  • MinerU 全文解析:将 PDF 规范化为 Markdown,每篇文献使用独立图片目录和相对链接。
  • Obsidian 文献库:自动维护 Literature/index.md、Literature/Literature.base、主笔记、PDF、全文和 Wiki。
  • 结构化研究层:支持 full_read、literature_review、passage_qa、figure_qa、concept 五类 Analysis。
  • 统一数据库视图:Literature/Analysis/Analysis.base 提供 9 个视图。
  • 科研 Skills:内置 paper-qa、full-read、passage-qa、figure-qa、compare-papers、literature-review、concept-learning。
  • 安全写入:支持 dry-run、staging、锁、备份、原子替换、事务预览和回滚。
  • 版本可验证:literature_version 返回当前版本、31 个工具、7 个 Skills 和五类 Analysis。
  • 多客户端接入:支持 Codex、Claude Code、OpenCode、Pi、Hermes 和 WorkBuddy。

效果展示

文献目录

<img src="https://raw.githubusercontent.com/luffysolution-svg/obsidian-vault-mcp/main/docs/assets/screenshots/v2/vault-structure.png" alt="Obsidian 文献目录" width="320">

Literature Index

<img src="https://raw.githubusercontent.com/luffysolution-svg/obsidian-vault-mcp/main/docs/assets/screenshots/v2/literature-index.png" alt="Literature Index" width="760">

多篇文献形成的可追溯 Wiki

<details> <summary>展开效果图</summary> <img src="https://raw.githubusercontent.com/luffysolution-svg/obsidian-vault-mcp/main/docs/assets/screenshots/v2/wiki-synthesis.png" alt="可追溯 Wiki 综合页面" width="780"> </details>

安装

3.0.2 已正式发布,要求 Python 3.10+。以下公开安装命令现已可用。

uv(推荐)

uv tool install "zotero-obsidian-mcp==3.0.2"
obsidian-vault-mcp --help

无需持久安装:

uvx --from "zotero-obsidian-mcp==3.0.2" obsidian-vault-mcp doctor --vault-path "<VAULT_PATH>"

pipx / pip

pipx install "zotero-obsidian-mcp==3.0.2"
# 或
python -m pip install "zotero-obsidian-mcp==3.0.2"

MCP Registry

io.github.luffysolution-svg/obsidian-vault-mcp

等价的 stdio 配置:

{
  "mcpServers": {
    "obsidian-literature": {
      "command": "uvx",
      "args": [
        "--from",
        "zotero-obsidian-mcp==3.0.2",
        "obsidian-vault-mcp",
        "serve",
        "--transport",
        "stdio"
      ],
      "env": {
        "OBSIDIAN_VAULT_PATH": "<VAULT_PATH>"
      }
    }
  }
}

首次配置

目标目录必须是已由 Obsidian 打开过的 Vault,并包含 .obsidian/。

obsidian-vault-mcp config init --vault-path "<VAULT_PATH>" --dry-run
obsidian-vault-mcp config init --vault-path "<VAULT_PATH>"
obsidian-vault-mcp config validate --vault-path "<VAULT_PATH>"
obsidian-vault-mcp doctor --vault-path "<VAULT_PATH>"
obsidian-vault-mcp call literature_version --json '{}'

启动 Zotero Desktop 并启用本地 API:

obsidian-vault-mcp call zotero_search_items --json '{"query":"photocatalysis"}'
obsidian-vault-mcp import item ABCD1234 --vault-path "<VAULT_PATH>" --dry-run
obsidian-vault-mcp import item ABCD1234 --vault-path "<VAULT_PATH>"

链接附件配置:

{
  "zotero": {
    "linkedAttachmentBaseDir": "<ZOTERO_LINKED_ATTACHMENT_BASE_DIR>"
  }
}

MinerU 解析:

obsidian-vault-mcp mineru parse ABCD1234 --vault-path "<VAULT_PATH>" --dry-run
obsidian-vault-mcp mineru parse ABCD1234 --vault-path "<VAULT_PATH>"

规范产物:

Literature/attachment/MinerU/ABCD1234.md
Literature/attachment/MinerU/image/ABCD1234/ABCD1234-fig01.png

Agent 与插件安装

obsidian-vault-mcp agent install codex --dry-run
obsidian-vault-mcp agent install codex

客户端名称可替换为 claude、opencode、pi、hermes 或 workbuddy。

客户端安装内容
Codex原生 marketplace 插件、MCP 和 7 Skills
Claude Code原生 marketplace 插件、MCP 和 7 Skills
OpenCode项目本地 MCP 和 7 Skills
Pi薄 TypeScript Extension
HermesMCP 配置
WorkBuddyMCP 配置

GitHub Release 中的离线插件包:

obsidian-vault-mcp-3.0.2-plugins.zip

Skills

Skill工作流
paper-qa单篇快速问答,默认不写入 Vault
full-read单篇完整精读并保存 full_read
passage-qa定位具体段落、方法、数据或结论
figure-qa解读图、表、Scheme 和方程
compare-papers对用户选定论文建立可比性矩阵
literature-review对文献池进行主题化综述
concept-learning跨文献建立概念模型

正式工具面

分组数量
版本、系统与配置5
Zotero6
导入与同步4
MinerU3
导航与校验3
Analysis5
Wiki3
事务2
合计31

发布一致性

3.0.2 必须同时出现在 Python 包、运行时 __version__、MCP Registry server.json、Codex/Claude 插件清单、Pi 包、Git Tag v3.0.2、GitHub Release 和 PyPI 中。Release workflow 会校验版本、Tag 和产物身份,构建 wheel、sdist、插件 ZIP,执行测试与 handshake,并生成 SHA256SUMS。

安全边界

  • 所有写操作先 dry-run,再提交并保存 transactionId。
  • 不要提交 Vault 绝对路径、Zotero 数据目录、MinerU token 或其他凭据。
  • MinerU 可能把 PDF 发送到外部服务,使用前确认授权和组织政策。
  • 推荐本地 stdio;SSE/HTTP 必须放在可信认证边界之后。
  • 事务备份不替代独立的 Vault 备份。

贡献者

感谢 方珸 / Lym Fang (@LimFang) 提出 Zotero 链接附件兼容方案。完整记录见 CONTRIBUTORS.md。

Related MCP servers

OMOmniScholar logo

Literature, Zotero, PDF, citation, materials, and scientific image tools for research agents.

4
Python
MIT
View repository →

MCP delegation fallback for AI agents to discover capabilities, knowledge, tools, and collaborators.

View repository →

Gerencie cobranças, invoices e atendimentos do RCM da Osigu diretamente do seu agente de IA.

View repository →

Read-only attack-path tools: reachable routes to sensitive assets, and what a fix would cut.

8
Go
Apache-2.0
View repository →
MAMaqui Analytics logo

Maqui Analytics

Maintained

Marketing analytics for local businesses — Instagram, ads, web traffic, SEO and reviews.

0
View repository →

Search and discover Lyra profiles.

0
TypeScript
MIT
View repository →