Sync Your Obsidian Vault with AI Memory

Obsidian is where you think. REM Labs is where AI remembers. This guide shows how to set up bidirectional sync between the two -- so your wikilinks become knowledge graph edges, your tags become searchable categories, and your entire vault becomes queryable by AI with 90% recall accuracy.

What the Sync Does

When you sync your Obsidian vault to REM Labs, each note is parsed and stored as a structured memory. The sync extracts:

• Title and content -- the full markdown body of each note
• Wikilinks -- every [[link]] in a note becomes an edge in the REM knowledge graph, connecting the two notes as related entities
• Tags -- #project, #meeting, #idea become searchable categories and metadata on the memory
• Frontmatter -- YAML frontmatter key-value pairs are preserved as structured metadata
• Vault path -- the file's location in your vault hierarchy is stored for context

Once synced, your notes are indexed three ways: vector embeddings for semantic search, full-text index for exact keyword matching, and entity graph for relationship traversal. You can search your vault by meaning, not just by filename or tag.

Option 1: Obsidian Plugin (Recommended)

1. Open Obsidian Settings, go to Community Plugins, and search for REM Labs. Alternatively, download from the GitHub repo.
2. Enable the plugin and paste your API key in the plugin settings. Get a free key at remlabs.ai/console.
3. Open the command palette (Ctrl/Cmd+P) and run REM: Sync Vault.

The plugin handles batching automatically -- vaults with more than 200 notes are split into batches. Unchanged notes are detected via SHA-256 content hashing and skipped, so repeated syncs are fast even for large vaults.

Option 2: API Import

If you prefer to script the sync or run it from CI, use the import endpoint directly.

The endpoint accepts up to 200 notes per request. Each note's title, content, path, tags, wikilinks, and frontmatter are extracted and stored as structured memories with full knowledge graph integration.

Export: REM Back to Obsidian

The sync is bidirectional. You can export memories from REM Labs as Obsidian-compatible markdown files with reconstructed frontmatter, tags, and wikilinks that drop straight into your vault.

This means memories stored by other integrations -- conversations imported from ChatGPT, context saved by the MCP server in Claude Code, or facts extracted from your Gmail -- can all be exported as Obsidian notes and incorporated into your vault.

Search Your Vault with AI

Once synced, you can search your vault through the REM Labs API using multi-signal fusion search.

This is not keyword search. The query "product launch timeline" will find notes about "shipping the SDK by June" and "Q2 delivery schedule" even if those notes never use the word "timeline." The four retrieval paths -- vector, full-text, entity graph, and metadata -- run in parallel and results are fused with neural reranking.

How Wikilinks Become Graph Edges

This is the part that makes the integration genuinely useful for Obsidian power users. When your note [[Launch Plan]] links to [[Q2 Goals]], REM creates a relationship edge between those two entities in its knowledge graph. When you later search for context about your launch plan, graph traversal can surface the Q2 Goals note as related context -- even if the semantic embedding of "Q2 Goals" is not particularly close to your query.

This mirrors how Obsidian's graph view works, but makes it queryable by AI. Your manually curated knowledge structure becomes part of the retrieval pipeline.

API Endpoints Reference