I would like to build a “second brain” with Hermes Agent and/or Claude and a database. Others use Obsidian. Is it possible to use immediately Devonthink in the same way? That would be great. I don’t like too much Obsidian and as I have already thousands and thousands of records in Devonthink, I imagine that this could be an interesting thing to melt everything together.
Yes- either one works with the DT4 MCP server.
Yes thank you for this confirmation. I already configured Devonthink access from Claude and also from AGY. That worked but I wanted to know what is the best way to create an efficient workflow with their intelligence, with Devon. The first thing I did is create a new database, which will be the playground for the AIs for writing logs of everything they did in the day and for putting all memory MD in it, at least as a copy.
other tips perhaps, please?
What are your overall goals that you want to achieve with Devonthink and/or with your AI agent(s)?
This is my PKIM structures built originally with the Swift AppleScript Bridge. It’s about a structured knowledge base, leveraging dt-links to allow you to separate knowledge from content.
I would like that evey knowledge and data i need and use is inside Devon and managed by modern AI. If even the skills and memories for all AI-agents (use Agy, Claude, and will install local Hermes while using it already on a remote server) are in the same place, that’s perfect.
that’s very interesting, will dig in it
This may be of interest:
Just let you know that for several months, I’m working with Hermes on my Devon databases and it works very fine. Hermes finds documents much faster than Devonthink does. And I let him do what he wants on my databases, have never been disappointed. Make new records, move records, look up documents, everything perfect.
Devon also has become the “brain” for Hermes, no need for Obsidian, Devon is much better and reallty universal.
I also replaced “DayOne” App by Hermes and DEvon: every thought that should be recorded as a diary record goes in a group “Diary”
Great…
Could you please explain that in more detail? In the end DEVONthink does the work in both cases.
A research with Hermes finds more precisely what i am looking for, it shows the most relevant things first, Devon doesn’t always.
I have noticed the same thing frequently. I suspect what is happening is two-fold:
(1) AI may be more knowledgable about the details of how to form a DT search than many/most DT users are.
(2) AI can parse a large set of results and using semantic reasoning to determine what is most applicable. The human user would have to read through all the results to reach similar conclusions.
This is not an apples-to-apples comparison. You’ve provided no information on your search terms in DEVONthink and your prompt to Hermes.
I think the point is true pretty generally.
It is correct that Devonthink and Hermes have the same capbility to initiate a standard Devonthink search → But AI often knows how to do that better than a typical DT user
And even for the most advanced Devonthink user - AI can often search semanticaly through a large response set and suggest the strongest hits a lot more efficiently than a human can read through it all
Overall that’s not a negative for Devonthink - it’s a huge plus as it greatly increases the scale of what Devonthink can do