Obsidian Brain - Global Agent Memory
A VS Code extension that provides global knowledge persistence for AI agents using an Obsidian vault, with workspace-scoped memory, integrated filesystem operations, and chat history search. Perfect for AI agents working across multiple projects.
Status: v2.4.0 - Production Ready (recommended for vaults <300 files)
Score: 78/100 - Solid v2.x product with room for enhancement
✨ Key Features
Store knowledge that persists across ALL workspaces and ALL projects forever.
- Instant Search - Indexed, scored search with
#tag support (zero-delay results)
- Structured Organization - Engineering/, Playbooks/, Decisions/, Snippets/ folders
- Quick Notes - Capture ideas to _Inbox with auto-timestamp
- Tag-Based Discovery - Find notes by
#topic hashtags
- Operations - search, list, read, write, append, quick_note, str_replace, insert, delete, move
Temporary workspace-specific context that helps during current work.
- Session Context - Track progress, decisions, blockers for THIS workspace
- Auto-Organized - Stored in
_Workspaces/[workspace]/
- Quick Access - view, create, str_replace, insert, delete, rename
- Perfect For - TODO lists, temporary notes, branch-specific context
File operations without external MCP dependency.
- Parallel Reads - read_multiple_files for batch operations
- Pattern Search - Glob-based file discovery
- Full Control - write_file, move_file, create_directory, get_file_info
📜 Chat History Tool
Never lose important conversations - search across ALL VS Code chat sessions.
- Full-Text Search - Find conversations about specific topics
- Date Filtering - Search "last week" or specific date ranges
- Export - Save valuable sessions to your Obsidian brain
- Statistics - View chat activity over time
🚀 Installation
- Install from the VS Code Marketplace
- Set your Obsidian vault path in VS Code settings
- Reload VS Code
- Start using the tools in Copilot Chat!
⚙️ Configuration
| Setting |
Description |
Default |
obsidianBrain.vaultPath |
Path to your Obsidian vault |
C:\workspace\obsidian-vault |
obsidianBrain.filesystemAllowedDirectories |
Additional directories for filesystem access |
[] |
💡 Usage Patterns
Session Start
Search brain for prior context on this topic
Check workspace memory for today's progress
Begin work with context awareness
quick_note: "Decided on JWT for auth because stateless"
quick_note: "Issue: CORS headers missing on prod"
quick_note: "Next: Add backoff logic to retry handler"
Session End
append to Engineering/[project].md with:
- What was accomplished
- Key decisions and why
- Problems and solutions
- Next steps
Long Sessions (50+ messages)
When context window is 75% full:
1. Summarize to brain: obsidian_brain: append
2. Update workspace memory: memory: create
3. Create handoff: Use handoff tool for new thread
| Vault Size |
Recommended |
Performance |
| <100 files |
✅ YES |
Instant (<100ms) |
| 100-300 files |
✅ GOOD |
Fast (<500ms) |
| 300-500 files |
⚠️ CAUTION |
Monitor performance |
| 500+ files |
❌ WAIT |
Unproven, contact for updates |
Note: Word index not cached - full rebuild on startup for large vaults.
🎯 Best For
- Solo developers building AI-assisted projects
- Multi-project workers needing cross-project knowledge
- Long coding sessions where context matters
- Decision documentation with ADR format in Decisions/
- Code snippets library in Snippets/
⚠️ Known Limitations
- No fuzzy search (typos will miss results)
- No backlinks/graph view (Obsidian app has this)
- No date range filtering
- Word index not cached (startup lag for 500+ files)
- VS Code only (use Obsidian app for mobile/other editors)
🔒 Security
- No encryption - Notes stored as plain markdown files
- No API keys needed - Direct filesystem access
- No external service - Everything local
- ⚠️ Don't store secrets - Use environment variables instead
📋 Commands
🧠 Obsidian Brain: Open Vault - Open vault in VS Code🧠 Obsidian Brain: Rebuild Index - Force reindex all files📜 Obsidian Brain: Search Chat History - Search past conversations📊 Obsidian Brain: Get Chat Stats - View chat activity
🏗️ Architecture
- VaultIndexer - In-memory index with file watching, cache persistence
- BrainTool - Markdown-based CRUD with instant indexing
- MemoryTool - Workspace-scoped storage (replaces Agent Memory)
- ContextWatcher - Token tracking for context overflow prevention
- FilesystemBackend - Direct VS Code filesystem API (no external MCP)
📈 Evaluation Results
After aggressive testing with 15+ subagent tests:
| Category |
Score |
Notes |
| Core Functionality |
17/20 |
Solid, instant indexing works perfectly |
| Performance |
15/20 |
Great for small/medium vaults |
| Real-World Usability |
16/25 |
Good core, missing fuzzy search |
| vs Alternatives |
22/25 |
Beats Agent Memory decisively |
| Deal Breakers |
8/10 |
No critical bugs, scale unknowns |
Verdict: ⚠️ CONDITIONAL RECOMMEND - Use for small-medium vaults with discipline.
🔄 Migration from Agent Memory
If you were using Agent Memory extension:
- Your workspace context now lives in
memory tool (same /memories/ behavior)
- Global knowledge moves to
obsidian_brain in your vault
- Key difference: Brain persists across workspaces (Agent Memory didn't)
- Much better search capabilities
- Requires Obsidian vault setup
🐛 Reporting Issues
Found a bug? Please report on GitHub Issues
Include:
- Vault size (approximate file count)
- Error message
- Steps to reproduce
📝 License
MIT
🤝 Contributing
Contributions welcome! This extension is actively maintained. Areas for improvement:
- Fuzzy search implementation
- Date range filtering
- Backlinks support
- Progress bar during indexing
🎓 Learning Resources
