Skip to content

The AI-native knowledge base for Goose

Goose is the open-source AI agent originally open-sourced by Block, now hosted by the Agentic AI Foundation. Hjarni is the knowledge base it reads across runs, machines, and the model you happen to be using that week.

Get started free See how it works

Free to start. No credit card required.

What this unlocks

Workflows that actually use your context

Hand Goose a brief that survives the run

Goose finishes a task and the working memory is gone. Save the project context, conventions, and gotchas to Hjarni once. Goose reads them at the start of the next run, no copy-paste.

Switch models without losing context

Goose lets you swap providers and models freely. Your Hjarni notes are model-agnostic. Whatever model is mounted today reads the same architecture decisions and runbooks.

Capture the lesson before the session ends

Ask Goose to write a Hjarni note describing the fix, the why, and the rollback. Tomorrow's run reads it back instead of rediscovering the dead ends.

One source of truth across CLI and chat

Goose reads from the terminal. Claude and ChatGPT read from the browser. The same Hjarni notes serve all three so your humans and your agents stop drifting apart.

Make Goose write back, not just read

Goose can call any MCP tool the server exposes. Hjarni's MCP tools include search, read, create, update, link, and tag. Useful for saving the outcome of a debugging session as a structured note.

Setup

Connect Goose in about two minutes

  1. 1

    Install Goose (see goose-docs.ai for the current command for your OS) and sign up for Hjarni.

  2. 2

    Run `goose configure`, choose Add Extension, then Remote Extension. Name it `hjarni` and paste the URL `https://hjarni.com/mcp`.

  3. 3

    On first connection Goose opens a browser tab for the Hjarni OAuth flow. Sign in with your Hjarni account and authorize.

  4. 4

    Start a session with `goose session` and ask the agent to list your Hjarni notes. If they appear, the extension is live.

  5. 5

    Optional: edit `~/.config/goose/config.yaml` directly if you prefer to keep extensions in version control. The snippet above is the shape Goose expects.

Add Hjarni as an extension via `goose configure` (Add Extension → Remote Extension)

extensions:
  hjarni:
    type: streamable_http
    uri: https://hjarni.com/mcp
    enabled: true

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 an open-source agent benefits from an external knowledge base

Goose is built around the agent loop: plan, act, observe, repeat. That loop is good at executing once it has facts to work with. Hjarni is where you write those facts down. Architecture decisions, runbooks, customer feedback, the conventions you would otherwise paste into every session.

Goose is open-source (Apache 2.0) and model-agnostic. Your provider is yours to pick. Hjarni stores plain Markdown notes you can export anytime. Swap CLI, swap model, swap host. The notes keep working. There is no proprietary memory format to migrate.

Goose treats MCP extensions as first-class. Extensions are the primary way Goose gains new capabilities. Hjarni is one HTTP endpoint plus OAuth, no local subprocess to babysit. Adding it to a Goose session is the same flow as adding any other remote extension.

Common questions

Questions before you connect Goose

Is Goose really open-source?

Yes. Goose was originally open-sourced by Block under the Apache 2.0 license and is now hosted by the Agentic AI Foundation at the Linux Foundation. You bring your own model provider and your own MCP extensions. Hjarni is one of those extensions.

Where does the Goose config file live?

At ~/.config/goose/config.yaml on macOS and Linux. Goose creates it on first run. You can also use `goose configure` from the terminal to add extensions interactively without editing YAML by hand.

Does Goose support remote (HTTP) MCP servers?

Yes. Goose calls them Remote Extensions and supports the Streamable HTTP transport that Hjarni exposes. Older builds bridged via stdio; current builds connect directly over HTTP, with OAuth handled in the browser on first use.

Can Goose create notes back into Hjarni?

Yes. Hjarni's MCP server exposes tools for search, read, create, update, link, and tag. Goose can call any of them, so you can ask the agent to save the outcome of a session as a Hjarni note instead of losing it when the run ends.

Will swapping models in Goose break the Hjarni connection?

No. The Hjarni connection is at the extension level, not the model level. You can swap from Claude to GPT to a local Ollama model and the Hjarni notes are still available to whichever model is mounted.

Give Goose a memory

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

Get started free See pricing

More integrations

Hjarni for Claude Hjarni for ChatGPT Hjarni for Cursor Hjarni for Claude Code Hjarni for Windsurf Hjarni for Cline Hjarni for GitHub Copilot Hjarni for Zed Hjarni for Continue.dev Hjarni for Aider Hjarni for Open WebUI Hjarni for Raycast Hjarni for Alfred Hjarni for Obsidian

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.