Setting Up MCP for Tana
MCP sounds great in theory. Let’s make it work in practice. We’ll set up both layers: Tana Local MCP as your foundation, then supertag-mcp lite as the enhancement. By the end you’ll have AI connected to your live workspace and your local analytics.
By the end of this lesson, you’ll be able to configure both MCP servers in Claude Desktop, Claude Code, or Cursor.
Setup Overview
We’ll configure in two steps:
- Step 1: Tana Local MCP — the foundation. Live workspace access, read/write, tags, fields. Start here.
- Step 2: supertag-mcp lite — the enhancement. Semantic search, analytics, offline. Add when ready.
Note: You can start with Step 1 alone and add Step 2 later. Both servers run side by side without conflicts.
Step 1: Tana Local MCP (Foundation)
Tana provides an official local MCP server built into the Tana desktop app. It connects AI assistants directly to your live workspace for reading, writing, searching, and managing nodes, tags, and fields in real time.
To set up Tana Local MCP, follow the official guide:
→ Tana Local API/MCP documentation
The key steps are:
- Open Tana Desktop → User Settings → Tana Labs
- Enable “Local API/MCP server”
- Configure your AI client to connect (the docs cover Claude Desktop, Claude Code, Cursor, and others)
- Restart your AI client
Verify Layer 1
Ask Claude any of these:
- “List my Tana workspaces”
- “Search my Tana for meetings”
- “What tags do I have in Tana?”
If it responds with real data from your workspace, Layer 1 is working.
Note: Tana Local MCP requires the Tana app to be running. If you close Tana, the tools become unavailable.
Step 2: supertag-mcp lite (Enhancement)
supertag-mcp lite adds analytics, semantic search, and offline capabilities on top of your foundation. In lite mode, it only exposes tools that complement Tana Local MCP — no overlap, no duplication.
Prerequisites
# Check supertag-cli is installed
supertag --version
# Should show 2.0.0 or higher
# Check you have an export
supertag stats
# Should show your node counts
# Check MCP server exists
which supertag-mcp
# Should show a path
If anything fails, run supertag-export run first to create or update your export.
Note: Supertag 2.0 is compatible with Tana’s local MCP server. Lite mode ensures the two servers complement rather than duplicate each other.
Claude Desktop
Add a supertag entry alongside your existing tana-local config:
{
"mcpServers": {
"supertag": {
"command": "supertag-mcp",
"args": ["--lite"]
}
}
}
Restart Claude Desktop. You should now see tools from both servers.
Claude Code
claude mcp add supertag "supertag-mcp" -- --lite
Restart Claude Code. Both tool sets appear at startup.
Cursor
Add a second MCP server in Settings → Features → MCP:
- Name:
supertag - Command:
supertag-mcp - Args:
--lite
Restart Cursor.
Verify Layer 2
With both servers running, try:
- “Search my Tana for ‘productivity’” → uses
tana_search(supertag lite) - “What are my workspace stats?” → uses
tana_stats(supertag lite) - “Find notes related to project management” → uses
tana_semantic_search(supertag lite, if embeddings configured)
When to Add supertag-mcp lite
Not everyone needs Layer 2 immediately. Here’s when it adds value:
| Scenario | Layer 1 Only? | Add Layer 2? |
|---|---|---|
| Read/write live Tana data | Yes | Optional |
| Full-text search across all nodes | — | Yes |
| Semantic search by meaning | — | Yes |
| Analytics and aggregations | — | Yes |
| Offline access (Tana closed) | — | Yes |
| Graph traversal and related nodes | — | Yes |
| Simple tag/field management | Yes | Optional |
Start with Layer 1. Add Layer 2 when you find yourself wanting search, analytics, or offline capabilities.
Troubleshooting
Tana Local MCP issues:
| Problem | Solution |
|---|---|
| ”MCP server not found” | Ensure the Tana app is installed and running |
| No workspaces returned | Open a workspace in Tana first — only loaded workspaces are visible |
| Tools stop working | Check if Tana app is still running — it must stay open |
| Can’t write/edit | Verify your Tana app version supports the local MCP server |
supertag-mcp lite issues:
| Problem | Solution |
|---|---|
| ”MCP server not found” | Run which supertag-mcp — ensure it’s in PATH |
| ”No tools available” | Restart the AI client completely |
| ”Database not found” | Run supertag-export run to create export |
| Tools return empty results | Check supertag stats — do you have data? |
| Duplicate tools with Tana Local | Use --lite flag to enable lite mode |
Both servers:
| Problem | Solution |
|---|---|
| AI confused about which tool to use | Be specific: “Use Tana Local to create…” or “Use supertag to search…” |
| Config changes don’t take effect | Always restart the AI client after config changes |
Practice
Exercise: Get Connected
- Set up Tana Local MCP (Layer 1) for your preferred client
- Restart and verify: “List my Tana workspaces”
- Set up supertag-mcp lite (Layer 2) alongside it
- Restart and verify: “Search my Tana for [topic you know exists]”
- Test both layers: “Create a todo in Tana called ‘MCP test’” (Layer 1) then “How many todos do I have?” (Layer 2)
Checkpoint
Both MCP layers are running in your AI client.
After adding an MCP server to your config, what’s the most important step people forget?
Answer: Restarting the AI client. Configuration changes don’t take effect until restart. This applies every time you change the config for either server.
Key Takeaways
- Start with Tana Local MCP — it’s the foundation for live workspace access
- Add supertag-mcp lite when you need search, analytics, or offline access
- Both servers run side by side in the same config file — no conflicts
- Configuration is JSON — always restart your client after changes
- Use
--liteflag for supertag-mcp to avoid tool duplication
What’s Next
Making your first real query — asking Claude about your notes using both tool layers.