Skip to content

The AI-native knowledge base for Raycast

Raycast is the launcher. Raycast AI is the assistant. Hjarni is the knowledge base it reads, one keystroke away.

Free to start. No credit card required.

What this unlocks

Workflows that actually use your context

Ask your notes without opening a browser

Hit the Raycast hotkey, ask the AI a question, and let it search Hjarni for the answer. No tab switching, no copy-paste, no re-explaining.

Capture a thought without breaking flow

Trigger Raycast AI and dictate the note. If your Raycast build surfaces Hjarni's write tools you can have it create the note in the right folder; if not, dictate and paste, and the result still lands in Hjarni in seconds.

Pull a snippet into any app

Ask Raycast AI to read the relevant Hjarni note and paste a summary into the email, doc, or chat you are writing. Frontmost-app aware recall.

Retire the half-stale snippet library

If your Raycast snippets are full of long paragraphs that drift, move the long-lived ones to Hjarni. Edit once, Raycast AI reads the live version every time.

One knowledge base across the desktop

The same Hjarni notes Raycast AI reads on your Mac are the ones Claude reads in the browser and Cursor reads in your editor. Switching tools never costs you the notes.

Setup

Connect Raycast in about two minutes

  1. 1

    Sign up for Hjarni and have Raycast AI enabled. MCP server support requires a Raycast Pro plan.

  2. 2

    Trigger Raycast and search for 'Install MCP Server'. If you already have other servers, 'Manage MCP Servers' → Install New Server gets you to the same form.

  3. 3

    Set the name to hjarni, paste https://hjarni.com/mcp as the server URL, pick HTTP as the transport, and leave the OAuth type on Dynamic so Raycast handles the sign-in.

  4. 4

    Raycast opens the Hjarni OAuth page in your browser. Sign in and approve.

  5. 5

    Trigger Raycast AI (cmd+space then type 'AI', or the @ mention in Quick AI) and ask it to search Hjarni for one of your notes.

Raycast → run 'Install MCP Server' (or 'Manage MCP Servers' → Install New Server)

Name:        hjarni
Server URL:  https://hjarni.com/mcp
Transport:   HTTP
OAuth Type:  Dynamic (Raycast handles the sign-in)

The Hjarni MCP endpoint is https://hjarni.com/mcp. The connection uses OAuth, so you sign in with your Hjarni account when prompted. No API key files to manage.

Why a Raycast MCP knowledge base beats per-app shortcuts

Raycast's value is collapsing a hundred tiny actions into one launcher. The matching move on the knowledge side is collapsing a hundred scattered notes apps into one knowledge base your launcher can read. Hjarni is that layer, and MCP is the connector.

Snippet libraries and clipboard history are great for short things. They are the wrong place for the longer notes — the customer interview summary, the deploy runbook, the past decision. Those belong in a real knowledge base. Hjarni keeps them, Raycast AI reads them.

Raycast is one of several places you ask questions during the day. The browser is another. The editor is a third. Hjarni gives all three the same answer because they all read the same Markdown notes through MCP.

Common questions

Questions before you connect Raycast

Do I need Raycast Pro?

Yes. Raycast's MCP server support is a Raycast Pro feature. The free Raycast tier does not currently expose the MCP install commands. Check Raycast's current plan matrix for the latest.

Where does the Raycast MCP config live?

Raycast manages installed MCP servers internally. Trigger the launcher and run 'Install MCP Server' to add one, or 'Manage MCP Servers' to edit or remove existing servers. There is no JSON file to hand-edit.

Does Hjarni work with Raycast snippets and quicklinks?

Indirectly. Raycast AI can read Hjarni notes through MCP, then you can pipe its output into a snippet or quicklink. The MCP integration is the connector, snippets and quicklinks are the delivery.

Can Raycast AI create notes back into Hjarni?

Hjarni's MCP tools include create, update, link, and tag. Whether Raycast AI surfaces them depends on Raycast's model, permission prompts, and the way you phrase the request, but when it does, the write lands as a normal Hjarni note.

Will this work with Raycast for Windows once it ships?

Hjarni's MCP endpoint is platform-agnostic — same URL, same OAuth. As long as the Windows build of Raycast keeps the same MCP install commands and HTTP transport, the same configuration applies.

Give Raycast a memory

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

Write once. You both remember.

Free to start. No credit card required.

Give your AI a memory

Works with Claude and ChatGPT today. Gemini coming soon.