Hjarni for LM Studio

The AI-native knowledge base for LM Studio

LM Studio runs open models on your own hardware. Hjarni is the memory those local models read: durable notes available at query time without leaving your machine for context.

Get started free See how it works

Free to start. No credit card required.

What this unlocks

Workflows that actually use your context

Context for an offline model

A local model in LM Studio has no memory and no web. Connect Hjarni and it reads your notes at query time, keeping the model and your hardware in control.

Notes that outlive the chat

Each LM Studio session starts fresh. Write conventions, runbooks, and decisions to Hjarni once and every later session reads them back.

Read and write from the model

Hjarni's MCP tools include search, read, create, update, link, and tag. Ask the model to capture a result as a note instead of losing it when you close the chat.

Same notes in the app and the server

LM Studio supports remote MCP servers in both its chat UI and its local API server, so the assistant you run interactively and the one behind your API read the same Hjarni notes.

Setup

Connect LM Studio in about two minutes

1
Sign up for Hjarni. LM Studio supports MCP from version 0.3.17 onward, so update if needed.
2
Open the Program tab in LM Studio and choose Edit mcp.json, or open ~/.lmstudio/mcp.json directly.
3
Add the hjarni entry above. Leaving out headers triggers the OAuth browser flow on first use.
4
If you prefer a token, add an Authorization Bearer header with a Hjarni personal API token instead of using OAuth.
5
Load a model that supports tool use and ask it to list your Hjarni notes to confirm the connection.

Add Hjarni to mcp.json (~/.lmstudio/mcp.json)

{
  "mcpServers": {
    "hjarni": {
      "url": "https://hjarni.com/mcp"
    }
  }
}

Edit mcp.json from the LM Studio Program tab, or directly at ~/.lmstudio/mcp.json (it follows Cursor's notation). With no headers, LM Studio opens a browser to authorize via OAuth and stores the token for you. To use a token instead, add an Authorization Bearer header with a personal API token from Hjarni settings.

Why running models locally still leaves a memory gap

LM Studio is built for keeping inference on your own machine. That gives you privacy and zero per-token cost, but the model still has no idea who you are or what you decided last week. Local does not mean it remembers.

Hjarni fills that gap with Markdown notes you own and can export anytime. The model stays local; only the notes it requests travel over the MCP connection, and only when you connect it.

Because LM Studio supports remote MCP with OAuth handled in the browser, setup is a short mcp.json entry. The same Hjarni notes are then available to LM Studio's chat and to its local API server.

Common questions

Questions before you connect LM Studio

Does LM Studio support MCP?

Yes, since version 0.3.17, including remote MCP servers. Add Hjarni in mcp.json.

How does authentication work?

Add the server with no headers and LM Studio opens a browser to authorize via OAuth, then stores the token for you. Or add an Authorization Bearer header with a Hjarni personal API token.

Where is LM Studio's mcp.json?

At ~/.lmstudio/mcp.json (on Windows %USERPROFILE%\.lmstudio\mcp.json), editable from the Program tab. It follows Cursor's mcp.json notation.

Does my local model read notes without going online for them?

The model runs locally and reads only the notes it requests over the MCP connection. Your model and hardware stay in control of what gets pulled in.

Is it free?

Hjarni's free tier includes MCP access with no token limits, and LM Studio is free to use.

Give LM Studio a memory

The session starts where the last one ended. Write notes once. LM Studio reads them across every conversation.

Get started free See pricing

Give your AI a memory

Write once. You both remember.