Nexus MCP for Obsidian
Turns your Obsidian vault into an MCP-enabled workspace with tools for reading/writing notes, managing folders, running semantic searches, and maintaining long-term memory—all while keeping data local to your vault.
README

Nexus MCP for Obsidian
Nexus turns your Obsidian vault into an MCP-enabled workspace. It exposes safe, structured tools that AI copilots can call to read/write notes, manage folders, run searches, and maintain long-term memory—all while keeping data local to your vault.
Nexus is the successor to Claudesidian. Legacy installs in
.obsidian/plugins/claudesidian-mcp/still work.
Highlights
- Two-Tool Architecture – Just 2 MCP tools (
getTools+useTools) replace 40+ individual tools, reducing upfront token cost by ~95%. - Native Chat View – Stream tool calls, branch conversations, and manage models directly inside Obsidian.
- Workspace Memory – Workspaces, states, and traces in
.nexus/(sync-friendly JSONL + local SQLite cache). - Local Semantic Search – Desktop-only embeddings via sqlite-vec vector search—no external API calls.
- Full Vault Operations – Create, read, update, delete notes, folders, frontmatter, and batch edits.
- Multi-Provider Support – Anthropic, OpenAI, Google, Groq, Mistral, OpenRouter, Perplexity, Requesty, plus local servers (Ollama, LM Studio).
- Multi-Vault Ready – Independent MCP instances per vault.
Platform Notes
| Feature | Desktop | Mobile |
|---|---|---|
| Native Chat | ✅ | ✅ |
| MCP Bridge (Claude Desktop) | ✅ | — |
| Local Providers (Ollama/LM Studio) | ✅ | — |
| Semantic Embeddings | ✅ | — |
| Cloud Providers | ✅ | ✅ (fetch-based only) |
Two-Tool Architecture
Nexus exposes exactly 2 tools to MCP clients like Claude Desktop:
| Tool | Purpose |
|---|---|
toolManager_getTools |
Discovery – Returns schemas for requested agents/tools |
toolManager_useTools |
Execution – Runs tools with unified context |
Why This Matters
- ~95% token reduction in upfront schemas (~15,000 → ~500 tokens)
- Works great with small context window models (local LLMs)
- Context-first design captures memory/goal for every operation
Context Schema
Every useTools call includes context that helps maintain continuity:
{
workspaceId: string; // Scope identifier (name or UUID)
sessionId: string; // Session name (system assigns standard ID)
memory: string; // Conversation essence (1-3 sentences)
goal: string; // Current objective (1-3 sentences)
constraints?: string; // Rules/limits (1-3 sentences, optional)
}
Available Agents & Tools (34 total)
| Agent | Purpose | Tools |
|---|---|---|
| contentManager | Note reading/editing | read, write, update |
| storageManager | File/folder management | list, createFolder, move, copy, archive, open |
| searchManager | Search operations | searchContent, searchDirectory, searchMemory |
| memoryManager | Workspace/state management | createWorkspace, listWorkspaces, loadWorkspace, updateWorkspace, archiveWorkspace, createState, listStates, loadState |
| promptManager | Custom prompts & LLM | listModels, executePrompts, listPrompts, getPrompt, createPrompt, updatePrompt, archivePrompt, generateImage, subagent |
| canvasManager | Canvas operations | read, write, update, list |
Install
- Download the latest release assets:
manifest.json,styles.css,main.js,connector.js - Place them in
.obsidian/plugins/nexus/(or keep legacy.obsidian/plugins/claudesidian-mcp/) - Enable Nexus in Obsidian Settings → Community Plugins
- Restart Obsidian after first install
Configure Claude Desktop
Add the Nexus server to your claude_desktop_config.json:
{
"mcpServers": {
"nexus-your-vault": {
"command": "node",
"args": [
"/path/to/Vault/.obsidian/plugins/nexus/connector.js"
]
}
}
}
Or use the one-click setup: Settings → Nexus → Get Started → MCP Integration → Add Nexus to Claude
After adding, fully quit and relaunch Claude Desktop.
Using Native Chat
- Configure a provider in Settings → Nexus → Providers
- Open chat via ribbon icon or command palette (Nexus: Open Nexus Chat)
- Type
/for tools,@for custom agents,[[to link notes - Tool calls stream live with collapsible result panels
Workspace Memory
All data lives in .nexus/ inside your vault:
.nexus/
├── conversations/*.jsonl # Chat history (syncs across devices)
├── workspaces/*.jsonl # Workspace events
└── cache.db # SQLite cache (auto-rebuilt, not synced)
- Each tool call is tagged to a workspace automatically via context
- Create/load workspaces and save immutable state snapshots via tools or chat UI
- Archive workspaces and states for cold storage (restorable)
- No external database required
Semantic Search
Use searchManager.searchContent with semantic: true for meaning-based search:
- Desktop only – Embeddings run locally via iframe-sandboxed transformers.js
- Model:
Xenova/all-MiniLM-L6-v2(384 dimensions, ~23MB, cached in browser IndexedDB) - Vectors: Stored in
.nexus/cache.dbvia sqlite-vec for fast similarity search - First run downloads the model (requires internet); subsequent runs are fully offline
- Watch the status bar for indexing progress; click to pause/resume
Multi-Vault Setup
- Each vault runs its own MCP server with key
nexus-[vault-name] - Add one entry per vault in your MCP client config
- Keep vaults open to keep their servers reachable
Security & Privacy
- MCP server binds locally only—no remote listeners
- All file operations stay inside the active vault
- Network calls only for remote LLM providers (per your API keys)
- Embeddings download once, then run fully on-device
Troubleshooting
| Issue | Solution |
|---|---|
| Server not found | Settings → Nexus → Get Started → MCP Integration → Add Nexus to Claude, then restart Claude Desktop |
| Pipes not created | Ensure Obsidian is open; Windows uses nexus_mcp_<vault> named pipes |
| WebLLM crashes | Currently disabled due to WebGPU bug on Apple Silicon; use Ollama or LM Studio |
| Legacy install | Paths to .obsidian/plugins/claudesidian-mcp/connector.js still work |
Development
npm install # Install dependencies
npm run dev # Development build with watch
npm run build # Production build
npm run test # Run tests
npm run lint # Run ESLint
See CLAUDE.md for architecture details and contribution notes.
License
MIT License - see LICENSE for details.
Questions or issues? Open a GitHub issue with your OS, Obsidian version, and any console logs.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.