I’m tempted to come back from Tana to Logseq because Claude desktop can read and work with my MD files and become my interface totalk with my notes and update them.
Any advice? Would it work better with DB and MCP than directly with the MD files?
2 Likes
I’ve moved from ChatGPT to Claude because of this.
The MCP is under development. Have Claude talk to the API directly. It works well though it has to figure out Logseq’s model. You’ll need to create an API key for it. I had Claude create a script for API access so the shape matches the allow patterns in settings.json.
Edit: add script
I can’t comment on the db version, but I setup a dedicated software-dev graph and let claude-code access the md files directly. I setup some skills so it does it consistently with a playbook for how to format pages, tags add properties. It is working partially as a cross-session memory for projects. The only issue I find is context gets eaten quickly consuming all the md. YMMV
I had some dificulties to use the DB version in integration with Claude Code/Codex/Copilot CLI’s, so I prefer using Logseq OG because of that as well, the work with .md becames easier when using this tools, that’s why also I started a repo with a skill set to maintain and improve graphs using the this tools, it might help you guys on work with it because it helps the seemless use within the folder: https://github.com/gustavo-meilus/logseq-rig