Colle vs Hjarni

When to choose Hjarni

Hjarni is a hosted Markdown knowledge base with a built-in MCP server, so Claude and ChatGPT can both read it and write to it. It is listed in the Claude Connectors Directory, and you can export your notes to Markdown.

On the Teams plan, at $20 per seat per month, everyone keeps their own ChatGPT or Claude, and the team shares folders, AI instructions and per-folder roles (viewer, editor, admin).

Choose Hjarni when the knowledge should stay put and hold true for everyone: runbooks, decisions, how your team writes a proposal, what a long-standing client prefers. If the question is “what do we know?”, a shared knowledge base answers it each time anyone asks, in their own assistant.

When to choose Colle

Choose Colle when a piece of work changes hands. You tell your assistant “send this to James, he runs the launch now”, and James gets a handoff with the summary, the key facts and the files. A Chrome pop-up reaches him in about 0.2 seconds, and Web Push works on desktop and phone without the extension. He accepts with one click and asks his own Claude or ChatGPT “what did I get?”.

From then on the handoff has an owner and a status: accepted, returned with a reason, done, or overdue when the deadline passes. You can recall one you sent to the wrong person. Deadlines, amounts and decisions are recorded as facts with their history, so nobody has to ask “what was the deadline again?”. You can send to up to five people at once, or to a space, where every member is notified and one of them picks the work up. The details are on how it works.

Colle is not a knowledge base, so it is not the best fit if what you mainly want is a library your AI reads from. Data is hosted in the EU (Frankfurt), and administrators can export a tamper-evident audit log; more on the security page. Colle costs €15 per active person per month, or €12.50 billed yearly, excluding VAT, and people who only receive notifications do not pay. See pricing.

Using both

The two tools do different jobs, so they sit together well: keep the team's lasting knowledge in Hjarni, and hand off the task that uses it with Colle. The onboarding runbook lives in Hjarni; “Give Tom the client brief and our meeting notes” goes through Colle, so Tom is notified, the deadline is recorded as a fact and you can see when he accepts.

Both work through the assistant each person already uses, so nobody has to switch tools. Setup for each assistant is on the Claude and ChatGPT pages.