About smart-connections
smart-connections is a read-only MCP server for searching an Obsidian vault with Smart Connections embeddings. It serves Claude Code and other MCP clients, with local text-query inference and strict path validation.
Key features
- 01Text and semantic note search
- 02Local query embedding inference
- 03Read notes and list indexed notes
- 04Read-only, path-validated access
- 05Offline operation
Use cases
- 01Find vault notes from a text question
- 02Locate notes similar to a given note
- 03Retrieve a note or inspect indexed notes
MCP support
Exposes tools to MCP clients for text, embedding, and similar-note searches, note retrieval, indexed-note listing, and model information.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @gogogadgetbytes/smart-connections-mcpVAULT_PATHAbsolute path to your Obsidian vault (e.g., /Users/you/Documents/MyVault)Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMESmart Connections MCP Server
A security-first MCP server for Smart Connections. Read-only. Path-validated. Auditable.
Exposes Smart Connections embeddings to Claude Code and other MCP clients for semantic search of your Obsidian vault.
Why This Exists
We needed semantic search of our Obsidian vault from Claude Code. Existing options have problems:
- No path validation - User input passed directly to file operations
- Write access - Some expose mutation tools we don't need
- Heavy dependencies - PyTorch/transformers for what's essentially vector math
This implementation:
- Minimal dependencies - MCP SDK + Transformers.js (for text search)
- Fail-closed security - path validation with realpath, symlink detection
- Auditable - small TypeScript codebase you can actually read
Features
- Text search - query with plain text, not just note paths
- Semantic search using Smart Connections embeddings
- Local inference - uses Transformers.js (same model as Smart Connections)
- Read-only - no write operations, no shell execution
- Secure - strict path validation, bounded responses
- Offline - works without Obsidian running
Security Model
| Property | Guarantee |
|---|---|
| Path confinement | All file access validated against vault root |
| No traversal | ../ and symlink attacks blocked |
| Read-only | No write operations exposed |
| Bounded responses | Capped results (50), content length (10KB) |
| Fail closed | Errors deny access, never bypass |
| Audit logging | Security events logged with context |
Installation
Prerequisites
- Node.js 18+
- Obsidian with Smart Connections plugin installed
- Embeddings built (open vault in Obsidian, let Smart Connections index)
Setup
git clone https://github.com/gogogadgetbytes/smart-connections-mcp
cd smart-connections-mcp
npm install
npm run build
Configure Claude Code
Add to your Claude Code config:
claude mcp add smart-connections \
-e VAULT_PATH="/path/to/your/obsidian/vault" \
-- node /path/to/smart-connections-mcp/dist/index.js
Or manually add to ~/.claude.json:
{
"mcpServers": {
"smart-connections": {
"command": "node",
"args": ["/path/to/smart-connections-mcp/dist/index.js"],
"env": {
"VAULT_PATH": "/path/to/your/obsidian/vault"
}
}
}
}
Restart Claude Code to load the server.
Usage
Once configured, Claude Code can use these tools:
Search by Text
"Search my vault for notes about backup strategies"
→ Uses search_by_text tool
Search Similar Notes
"Find notes similar to Topics/Claude_Code.md"
→ Uses search_similar tool
Get Note Content
"Show me the content of Topics/Obsidian.md"
→ Uses get_note tool
List Indexed Notes
"What notes are indexed in my vault?"
→ Uses list_indexed tool
Tools
| Tool | Description |
|---|---|
search_by_text | Search using freeform text (computes embedding locally) |
search_similar | Find notes semantically similar to a given note |
search_by_embedding | Search using a raw embedding vector |
get_note | Get content of a specific note (path validated) |
get_model_info | Get embedding model configuration |
list_indexed | List all indexed notes |
Configuration
| Variable | Required | Description |
|---|---|---|
VAULT_PATH | Yes | Absolute path to Obsidian vault |
Limitations
- Single vault - Configure one vault per MCP server instance
- Index from Smart Connections - Note embeddings come from the plugin; text queries are embedded locally
- No write access - By design; use Obsidian for edits
- First run downloads model - ~50MB model cached in
~/.cache/huggingface/
Development
# Build
npm run build
# Test with MCP Inspector
npx @modelcontextprotocol/inspector node dist/index.js
Contributing
See CONTRIBUTING.md. Security-focused PRs welcome.
Security
To report security vulnerabilities, please email [email protected]. Do not open public issues for security concerns.
License
MIT - see LICENSE
Credits
- Smart Connections by Brian Petro
- Model Context Protocol by Anthropic
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.