DEVONagent 4 Wishlist

I love the idea of DEVONagent, and I’d love to see a modern take on a personal search engine. My perhaps not so humble wish list:

  • Plugins for “corporate” knowledge bases, namely Confluence and Jira
  • A unified, modern design. It’d be great if DEVONagent looked just as much at home on macOS as DEVONthink does.
  • Responsive UI that keeps up with the keyboard.
  • Even tighter integration with macOS (Shortcuts actions?)

I’m curious to hear what other DA users have on their v4 wishlist.


Thank you for the feedback!

While future versions should be able to use more search engines & databases, the app will still only ship with plug-ins for frequently requested ones (as there are way too many out there to support & maintain all of them).

That’s definitely planned.

When exactly do you experience this?


I only have one big wishlist item, which is to change the ghastly grey text in the Digest view (especially) to a readable black. I suspect this is already planned; the longstanding thread on this was suddenly closed earlier this year. At any rate it’s the one thing that really hits my workflow; it makes the digest so unusable in DA itself that I’m reduced to exporting it to RTF and reformatting it with a Nisus macro.

In terms of functionality, it would be great to have more user-friendly tools for custom plugin creation than manually editing raw XML. I don’t know how important this is to most users, and the tutorial is very helpful if you do want to go there, but it’s probably the feature with the most untapped power sitting there waiting to be unlocked. The process generally involves a lot of trial and error, and while I don’t particularly mind tweaking fields in a text file – you learn an enormous amount about search optimisation along the way – I could see a more graphical approach streamlining and speeding the operation as well as making it a bit more inviting to the user.

1 Like

This is indeed under consideration but the truth of the matter is that creating plugins isn’t simple since there isn’t a standard approach to how sites and search engines are built. This is why there isn’t a one-size-fits-all plugin and each site has to be individually developed for.

This is only one of those foolish ideas I have: offer an API to create plugins. As easy as return a list of items containing text (html for richness, perhaps), URL or file name, and nothing more. This could even be useful to search into other things than internet.

1 Like

Yes, I can completely see that. As anyone who’s had a go at building their own knows only too well, every site and resource has its own quirks, and you can sometimes find yourself tweaking parameter after parameter in vain. Rather than some kind of low-code Lego for commoner configurations, I was primarily thinking of something (comparatively) modest like a GUI front end to the .plist parameter editing. The current text-based version isn’t difficult and I don’t mind it myself (as a 90s throwback who still jumps aboard handcoding opportunities where available), but my kids who never read manuals take one look and just think word-of-the-month Nope.

This is the weird thing about DEVONagent, really. Everyone runs web searches every day of their lives and DA is hands-down the best tool ever devised for this. But even on this forum pre-loaded with DEVONheads, the traffic is minute compared to the bustling community of DTland. Obviously people (myself included) spend their lives in DEVONthink, launching it on startup and living at least partly in it all through the working day, whereas DA is something you fire up for particular tasks. But there’s also something about DA, its never-quite-intuitive-as-it-feels-it-should-be UI, and (above all) the way users have been passively trained by the search-engine business model towards low-effort, low-quality, zero-apparent-cost implementations of internet technology’s single most powerful affordance, that collectively serves as a behavioural deterrent against investing even modest cognitive and budgetary capital in prising the power of internet search away from the systems that have arisen to distort and monetise it at the expense of actual informational quality. As the OP from which this thread has unravelled noted, this is finally now being challenged in ways that DA has been out in front of from the start, and there’s an opportunity here for the world to catch up. But it will need a user base who understand, and care enough about, what internet search actually involves and does – and how we can exploit it, rather than vice-versa.



And yes, DEVONagent is a very powerful but oft-misunderstood application.
Generally, people often come at it as it’s a front-end to Google and wonder why they didn’t get 1,320,443 results from their search for bacon recipes :stuck_out_tongue:
They see the filtered results and think, “Google does way better than this!!”

The default search sets are very useful in general searching, still producing good but filtered results. However, yes creating a bespoke plugin requires more thought and fiddling. I’m sure @cgrunenberg has ideas how to make it friendlier but I think there will always be some level of tweaking involved.


Unless hell freezes over and there will be a common standard for searching :grin:



Like Xcode?

1 Like

Ha! Perhaps with a slightly smaller footprint…

When navigating search results in the menu bar app that’s bundled with DA pro. The rest of the app is very responsive.

It’s great to hear that there are updates planed. I’m looking forward to seeing what comes out of it. We gripe because we love :slight_smile: .

I’d love to see Confluence and Jira integration. They require logins, but almost every technical team I’ve ever worked with has used both of them. I imagine folks outside of the tech industry might not hear about them much, but I’ve used both for work for many years.

And is the menu extra still searching while navigating?

No, I think it’s generating thumbnails. I just tried again and couldn’t get it to react the same as when I posted above, so… maybe it was a transient issue. :man_shrugging:t2:

I’ve had licences for DA for almost as long as DT, i.e., 10+ years. I’ve spent a few dozens of hours with it, had regularly used it for a while working on a long-term project. But know, it’s mostly lying dormant. I’ve seen and still see the breadth and depth of things you could do with it. Yet, I haven’t been able to make use of DA’s potential. I don’t think it’s necessarily the “behavioural deterrent against investing even modest cognitive and budgetary capital” that @NickLowe has mentioned. But then, maybe my definition of modest differs or my “cognitive” is too sub-modest. I’ve often wondered, i.e. whenever I see DA’s icon or its name being mentioned, what it would take for me to use it more frequently and to my informational gains.
a) Learn by do & copy. When you jump into programming or learning natural languages, you don’t start with manuals and grammar book, but right dive into it. IDEs come with demo projects, and that’s something that be helpful here as well. How do other people use it day by day, how does their UI look like. Demo “projects” with lots of examples right in the UI could showcase how pros use it and provide good starting points right in the UI. (Yes, I’ve seen lots of vids on DA, spent hours in this forum, might even have bought and read that book…)
b) Easy, frequent use cases. Part of the learning curve of DA is its Sysiphonian shape. My rock of DA knowledge always rolls down again after the downward motivation-to-learn curve crosses the only mildly upward benefits-from-use curve. Hence, my learning curve has been more oscillating below the threshold of usefulness and never reached that more a less continuous upward shape or a high-enough plateau. So, to avoid a steep learning curve, easy use cases in an easy UI might help.


DT introduced automatic time-triggered locking for old, inactive threads. The rationale behind this was that those threads more often than not appy to prior (major) versions of the respective software, and that the age of the thread was not always immediately apparent to the reader or poster. I would guess the thread in question was affected by that rule, rather than being actively closed.

I thought the time-triggered auto-close only kicked in three years after the last reply (mine, in this case); this one was only five months. But it’s true that the penultimate post had been in 2015 and I was deliberately bumping it, so maybe the algorithm is more nuanced than it’s letting on. (I haven’t moderated a Discourse forum for a couple of years now, so if I ever knew how this feature worked the memory has auto-closed itself.) The grey text is still as horrible a feature now as it was in 2015, though!

Smiling when i read this, elegantly said and precisely describes my experience with DA.

My heavy use of DA was many years ago when a few projects required something better than Google. Most around me in the business unit had PCs and hence no access to Devon technologies. With DA as secret weapon, I quickly became the search king who could find anything. Additionally building a competitor database on anything new they produced with robot like precision gained me many admirers. DA and the combination with DT was amazing to use.

I say “was” because i have little appetite nor time to relearn how this all works again. I do have a need for DA every now and then, fire it up and lustfully look at its interface knowing the power i need resides somewhere inside of it. But many websites are quite different now (a decade or so later) and getting to what i want is either not possible with current DA or requires more active knowledge of DA then i have now.

I am very happy to hear you may be working on improvements for DA. My hope is that this includes a more intuitive workflow for setting up ones own queries and (!) an API for a complete custom plugin.

Anyhow, i remain a fan of this software and strongly believe it fills a need many have yet to acknowledge.


When I bought it, I actually thought that DevonAgent would search my DevonThink pro database and use its artificial intelligence to make connections to text passages in documents. That would be a great thing, wouldn’t it?

Maybe it would also be possible to weight certain search terms heavier or lighter? It would also be great if you could influence the found connections with different parameters. These settings should be clearly visible graphically. Similar to Obsidian. In Obsidian are tons of ideas there on how to graphically represent connections.

Obsidian is probably the most popular software To extend one’s thinking electronically, keyword “Second Brain”. Maybe a link to the open source software would be useful?

I have to finish my studies shortly and wonder if I can use the update in the near future. Or is the release of the update for DevonAgent rather not this year?:question::bangbang:

Artificial intelligences, like Jaspers’, offer the ability to ask questions. Strangely, this seems to be a windy and expensive market. Still, I’ve already received meaningful answers to questions that have kept me going. Will there be links to artificial intelligences on in the new update? Would be nice if your company also provides a small plugin that can be used to include an artificial intelligence.