Obsidian Search Tool (Level 4)

Input

• Query string (e.g., "machine learning notes")

Output

• Collection ID containing one structured Note per search result
• Each Note contains JSON with uniform structure:

{
  "text": "# Machine Learning\n\nThis note discusses...",
  "format": "markdown",
  "metadata": {
    "source_url": "obsidian://vault/notes/machine-learning.md",
    "uri": "obsidian://vault/notes/machine-learning.md",
    "domain": "obsidian",
    "elapsed_ms": 45
  },
  "char_count": 1123
}

Configuration

Requires Obsidian Local REST API plugin to be installed and enabled in Obsidian.

• OBSIDIAN_MCP_URL environment variable (defaults to http://127.0.0.1:27123 for HTTP, or https://127.0.0.1:27124 for HTTPS)
• OBSIDIAN_MCP_API_KEY environment variable (required) - Get your API key from Obsidian Local REST API plugin settings

Search Behavior

Searches by filename first, then fetches note content and filters by content match. Only searches markdown files (.md).

Example Note Structure

Each Note in the returned Collection contains:

{
  "text": "# Note Title\n\nNote content here...",
  "format": "markdown",
  "metadata": {
    "source_url": "obsidian://vault/path/to/note.md",
    "uri": "obsidian://vault/path/to/note.md",
    "domain": "obsidian",
    "elapsed_ms": 50
  },
  "char_count": 523
}

Note: The uri field is a standardized URI field (alias for source_url) for consistency with search-web and search primitives. Use metadata.uri in project operations for consistent access across all tools.

Important: The value parameter must be a literal string (e.g., "grobid"), not a variable reference (e.g., "$grobid").

Common Workflows

Search and summarize:

{"type":"search-obsidian","value":"neural networks","out":"$notes"}
{"type":"summarize","target":"$notes","focus":"what are neural networks","out":"$summary"}

Filter results:

{"type":"search-obsidian","value":"python scripts","out":"$notes"}
{"type":"filter-collection","target":"$notes","predicate":"contains code","out":"$code_notes"}