Sound familiar?
Your AI starts from zero every conversation. You explain your stack, your customers, your conventions, every time you open a new chat.
Your notes were designed for you to reread. Not for your AI to read. Proprietary blocks, hidden databases, page builders. None of it survives the trip into a prompt.
The "AI feature" your tool ships does not see your structure, follow your rules, or write notes back.
What AI-native means in practice
A built-in MCP server, not a plugin. The product was scaffolded around the protocol. Connect Claude, ChatGPT, Cursor, or Claude Code in about a minute. They read your notes, search the folders you connect, follow your rules, and write back.
Per-folder instructions. Tell your AI how to write into each space. "When asked about deploys, search this folder first and follow the steps." "Quote interviewees verbatim." The rules ride along with the data, so every conversation starts with the right context.
Your agent works while you sleep. Save a rough note from your phone. Ask Claude to find the related decisions and link them. Wake up to a folder of linked notes that the AI wrote in your voice.
What this looks like
A folder structure. An instruction. An exchange.
How do we deploy the worker service?
From "Worker service deploy" in Runbooks > Deploys: "Stop the worker pod, push the new image tag, then run the post-deploy checks." Rollback is at the bottom of the note if step three fails.
Your folders. Your rules. Your AI reading them.
What this is not
Not another notes app with a chat bubble. Not a vector database you have to feed. Not a wrapper around an LLM. Hjarni is a place to write Markdown, organize it into folders, and tag it. The AI part is the MCP server that exposes all of that to the model you already use.
We left features out on purpose. No databases. No kanban. No page builders. The constraint is the feature. Your AI can only read what your notes app is willing to expose.