Why look past Obsidian for AI
Obsidian is an excellent local-first note tool. The friction shows up specifically around AI. Its MCP support comes from third-party plugins you wire up yourself, and because the vault lives on your machine, a hosted assistant like ChatGPT on the web has no clean way to reach it. You end up copying notes into the chat, which is the exact problem a knowledge base should remove.
So the question is rarely "is Obsidian good." It is "what is the least disruptive way to make my notes readable by my AI, without giving up Markdown."
Keep Markdown, add MCP
Hjarni is built for that move. Notes are plain Markdown, organized in folders with wiki-links, and the MCP server is built in rather than a plugin. You import your Obsidian vault as a ZIP and folders, wiki-links, frontmatter, and attachments come across. From then on Claude and ChatGPT read and write the same notes, and you can still export everything back to Markdown anytime.
Obsidian assembles your workflow from plugins. Hjarni connects your knowledge to whichever AI you ask.
What to look for
If AI memory is the goal, weigh three things: does it store plain Markdown you can export, is the MCP server built in or a plugin you maintain, and can both ChatGPT and Claude reach it from any device. Local-first tools score well on ownership but poorly on remote reach. Hjarni trades local storage for a hosted endpoint, which is what makes the cross-assistant part work.
When to stay on Obsidian
If you love local files, the plugin ecosystem, the graph view, or working fully offline, Obsidian is hard to beat and you should keep it. Plenty of people keep Obsidian for personal writing and use a hosted knowledge base for the notes they want their AI and team to read.