Hjarni for LibreChat

The AI-native knowledge base for LibreChat

LibreChat puts every model provider behind one self-hosted chat. Hjarni is the shared memory all of them read, so the same context follows you from GPT to Claude to your local model.

Get started free See how it works

Free to start. No credit card required.

What this unlocks

Workflows that actually use your context

One memory across every provider

LibreChat routes you between OpenAI, Anthropic, Google, and local models. Hjarni notes are model-agnostic, so whichever model answers reads the same architecture decisions, runbooks, and preferences.

Per-user context in a multi-user deployment

LibreChat supports OAuth for remote MCP, so each member signs in with their own Hjarni account on first use. Everyone gets their own notes, not a shared key.

Write back, not just read

Hjarni's MCP tools cover search, read, create, update, link, and tag. Ask a model in LibreChat to save the outcome of a conversation as a structured note for next time.

Stop re-explaining across model switches

Switching models mid-task usually means re-pasting context. With Hjarni connected, the new model picks up the same notes the last one was working from.

Setup

Connect LibreChat in about two minutes

1
Sign up for Hjarni and have your self-hosted LibreChat instance running.
2
Add the mcpServers block above to your librechat.yaml (or add the server from the LibreChat UI), using the streamable-http transport.
3
On first use LibreChat prompts each user to authorize via OAuth. Alternatively, create a personal API token in Hjarni settings and pass it as an Authorization Bearer header.
4
Restart LibreChat so it picks up the config, then enable the Hjarni tools for a model that supports tool calling.
5
Ask the model to list your Hjarni notes. If they appear, the connection is live.

Add Hjarni under mcpServers in librechat.yaml

mcpServers:
  hjarni:
    type: streamable-http
    url: https://hjarni.com/mcp

LibreChat supports OAuth 2.0 (Authorization Code with PKCE) for remote MCP servers, so each user authenticates with their own Hjarni account on first use. You can also use a personal API token as an Authorization Bearer header. The librechat.yaml schema evolves, so check the LibreChat MCP docs for the current field names.

Why a multi-model hub needs one external memory

LibreChat's whole premise is that you should not be locked to one model vendor. The catch is that each model starts from nothing: no shared memory of your project, your conventions, or what you decided last week. Hjarni is where that memory lives, as plain Markdown notes you own.

Because LibreChat is open-source and self-hosted, the notes stay yours on both ends. Hjarni stores them in the EU and exports the whole knowledge base as a Markdown ZIP whenever you want. Swap providers, swap models, and the notes keep working.

LibreChat treats MCP servers as first-class and supports remote ones over Streamable HTTP, so Hjarni is one endpoint plus OAuth, with no local subprocess to manage. In a team deployment, per-user OAuth means everyone reads and writes their own notes.

Common questions

Questions before you connect LibreChat

Does LibreChat support remote MCP servers?

Yes. LibreChat supports remote MCP over Streamable HTTP and SSE, and Hjarni uses Streamable HTTP. Add it under mcpServers in librechat.yaml, or from the LibreChat UI.

How do users authenticate to Hjarni in LibreChat?

LibreChat supports OAuth 2.0 with PKCE for remote MCP, so each member signs in with their own Hjarni account on first use. A personal API token passed as an Authorization Bearer header also works.

Do all my LibreChat users share one Hjarni account?

No. With OAuth, each user authorizes their own Hjarni account, so everyone reads and writes their own notes rather than a shared key.

Which models can use the Hjarni tools in LibreChat?

Any model you configure that supports tool calling. The notes are model-agnostic, so the same context serves OpenAI, Anthropic, Google, and local models behind LibreChat.

Is the Hjarni MCP integration free?

Hjarni's free tier includes MCP access with no token limits. LibreChat is open-source and self-hosted, so there is no extra Hjarni charge for connecting it.

Give LibreChat a memory

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

Get started free See pricing

Give your AI a memory

Write once. You both remember.