Obsidian Search

Semantic search for Obsidian notes using AI vector embeddings. Find notes by meaning, discover connections, filter by tags/dates/folders, and retrieve full content. Use when the user says 'search my notes', 'find in Obsidian', 'what did I write about', 'related notes', 'Obsidian search', or wants AI-powered knowledge retrieval from their Obsidian vault.

AI-powered semantic search for your Obsidian notes — find notes by meaning, not just keywords. Powered by Obvec.

Security & Data Handling

• MCP link is a credential: Your MCP Server URL (https://rag.10xboost.org/mcp?token=xxxxx) contains an embedded authentication token. Treat it like a password — do not share it publicly.
• Token scope: The token grants read-only access. It can search, list, retrieve note content, and analyze connections. It cannot modify, delete, or create notes in your Obsidian vault.
• Token lifetime: Tokens are valid for 30 days. After expiry, you will need to get a new MCP link from obsidian.10xboost.org. You can also regenerate the token at any time in Settings, which invalidates any previously shared links.
• Where the token is stored: The token is generated server-side by Obvec. When you paste the MCP link into a Claude Connector or MCP client config, that client stores the link locally on your device. Revoking or regenerating the token server-side invalidates all copies.
• Your note data: To provide semantic search, your Obsidian note content is uploaded to the Obvec server and stored as text and vector embeddings. This means your note content is stored on a third-party server (Obvec, hosted on Google Cloud). Notes are isolated per user and are not shared with other users or used for model training. You can delete your indexed data at any time from your account dashboard.
• No additional credentials: No separate API keys, environment variables, or secrets are needed beyond the MCP link.

Prerequisites

1. Sign up at obsidian.10xboost.org with Google
2. Connect your Obsidian vault — follow the setup guide to sync your notes
3. Get your MCP link: Go to Settings → copy your MCP Server URL
4. Add to Claude: Paste the MCP link as a Connector — no install, no API key needed

Available Tools (4)

Tool	Description
search_notes	Semantic search — find notes by meaning with similarity scoring, tag/date filters
list_notes	Browse notes by folder, tags, date range, with sorting options
get_note	Retrieve full content of a specific note by path or search term
analyze_connections	Discover related notes through AI-powered similarity analysis

Workflow

Step 1: Understand the User's Intent

User Request	Tool to Use
"Find notes about X"	search_notes
"What did I write about X?"	search_notes
"Show my recent notes"	list_notes with sortBy: "modifiedAt"
"Notes in folder X"	list_notes with pathPrefix
"Open note X"	get_note
"What's related to X?"	analyze_connections

Step 2: Search Notes

Semantic Search

search_notes(
  query: "machine learning project ideas",
  limit: 10,
  minScore: 0.7
)

With Filters

search_notes(
  query: "meeting notes",
  tags: ["work", "project-x"],
  sortBy: "modifiedAt",
  limit: 20
)

Step 3: Retrieve Full Content

get_note(path: "Projects/AI Research.md")

or search by term:

get_note(searchTerm: "quarterly review")

Step 4: Discover Connections

analyze_connections(
  reference: "Projects/AI Research.md",
  limit: 10,
  minScore: 0.6
)

Returns notes most semantically similar to the reference — great for finding related ideas, building knowledge graphs, or discovering forgotten notes.

Step 5: Present Results

• Search results: Show titles, relevance scores, and brief excerpts
• Note content: Display the full markdown content
• Connections: Present as a ranked list with similarity scores and note titles

Error Handling

Error	Solution
No results found	Try a broader query or lower minScore
Note not found	Use list_notes to find the correct path
Token expired	Get a new MCP link from obsidian.10xboost.org Settings (tokens last 30 days)