A clear and practical guide demonstrating how a multi-layered agentic assistant like Alfred can transform personal productivity, task management, and presence through secure, automated, and durable workflows
Hey David, I just found you (Alfred first actually) searching for a way to get my Zo instance connected to my Tailscale mesh. TLDR, I currently have my rudimentary agent in an OpenCode shell local on my mesh, but I'm planning to migrate him over to both OpenClaw and Agent Zero.
Questions:
Would Alfred be able to integrate with Agent Zero?
If not, can you add that integration in the future?
Congratulations on an extraordinary project. I'm very impressed and looking forward to future updates.BTW, please do more videos... maybe you could have Alfred create Youtube videos from the blog transcript!
Thank you very much for this breakdown! I’ve been following your posts for a while now because an "Alfred" is something I’ve wanted to build for a long time. I have OpenClaw running on my Raspberry Pi, which is where my agent—also named Alfred—lives. (Coincidence, I swear! I was thinking of Batman while onboarding OpenClaw).
Would it be possible to install your Alfred there as well? Python is already installed, but I’m not quite sure how to connect my OpenClaw instance to your Alfred running on my Pi. That was the one part I didn't quite get from your walkthrough.
yeah in theory is should install on your RPi just fine, just follow the same steps. the pip installer should find openclaw on PATH. if it doesn’t ,let me know. also be mindful that by default Alfred’s Surveyor uses local embeddings (nomic-embed-text) and some ML libraries (HBDSCAN, Leiden algorithm) that may or may not be too much for your RPi. either way let me know how it goes!
Thanks so much for sharing this. One thing my ADHD brain is grappling with is where to focus attention among all my subscription channels (Substack, podcast subscriptions, newsletters, articles shared by friends/family, etc.). Have you played at all with using Alfred to filter these channels and make suggestions (i.e curate your own "algorithm")?
I haven't built anything yet, but one idea that we discussed with my wife was to satiate my hunger for knowledge and my curiosity in a way that doesn't find me the screens would be some sort of automated creation feed that curates stuff I would want to read every week. But instead of sending it to my inbox it would automatically send it to a printing service who would then send it to me in some ways so I could get my own weekly magazine fully personalised for my interests all managed by Alfred.
I have no idea if it's possible, but that would be cool
That’s EXACTLY how I’ve been thinking of this project.
Douglas Rushkoff points out in his brilliant polemic, Team Human, that normal people are always one communications revolution behind the elite. E.g. priests of antiquity used their literacy to interpret the Bible to an illiterate set of commoners. But when the Guttenberg press was invented and commoners became literate, elite now controlled what books got published. It goes on like that through the telegraph, radio, television, the Internet, and now we’re seeing it beginning for social media, and because things are moving so fast, maybe even for agentic AI as well.
Beginning right here, in this comments section.
As those with the most systemic power now have access to build AI models and influence the world through their weights, we have an ability to build software. This is why we’re hearing about the death of SaaS right?
But it could (must?) also be the death of the dominance of the previous media revolution.
A good intro to how algorithms work might be the Netflix Doc (based on Rushkoff’s work) The Social Dilemma. If you want to dig deeper, Surveillance Capitalism by Shoshana Zuboff is maybe a little outdated now, but an excellent explanation of how the data that we generate by existing gets converted into advertising dollars by companies that can build a predictive profile of us.
And now, I’m seeing all these OpenClaw users naturally inclined to connect as much of their personal data to these tools as possible so that they can make predictions.
Well that’s LITERALLY democratizing the behavioral prediction algorithms and not to get grandiose, but there could be a massive boost in wellness and agency amongst the broader population if tools like Alfred handed that power to normal people.
Meanwhile, the tools themselves “go meta” on the previous communications revolution by bundling feeds and making a single umbrella under which the old world is contained. Just exactly as you suggested.
Love this. I'm actually building something similar, a marketing automation agent for indie devs that handles App Store optimization, Reddit launches, and content scheduling. The hardest part isn't the AI logic, it's defining the guardrails so the agent doesn't go rogue on your behalf.
Great writeup! I'm trying ZO now. You may want to check this out - Obsidian has a new beta headless version to sync from the cli, no need for the desktop version: https://obsidian.md/help/headless
I love how you architected Alfred. It is based on how humans think and do the work without AI, then how humans translate that work with the use of AI tools and any behavior change we, humans must make to truly make AI work for us. I will be attempting to install this! :-)
This is great workflow but Omi presents significant privacy risks as it is uncontrolled recording and storage in cloud. Is there an alternative setup you would recommend?
omi is a wearable microphone that's fully configurable. the default configuration is what you said, which i disabled.
my configuration is:
-> double button tap turns it on/off
→ default on at home, default off everywhere else
- omi service disabled, all recordings go to my own VPS → transcribed → processed → transcription and audio deleted
that said you dont need ambient recording. I do. sending voice notes does the same job. just record a voice recording on your phone and send it to the inbox folder.
but that assumes more clicks and button taps and a conscious decision to log which for me kills the purpose
Ok, I did everything you said and it's all running. Do I interact w Alfred through the openclaw telegram interface or do i need to set up something separate?
A clear and practical guide demonstrating how a multi-layered agentic assistant like Alfred can transform personal productivity, task management, and presence through secure, automated, and durable workflows
Hey David, I just found you (Alfred first actually) searching for a way to get my Zo instance connected to my Tailscale mesh. TLDR, I currently have my rudimentary agent in an OpenCode shell local on my mesh, but I'm planning to migrate him over to both OpenClaw and Agent Zero.
Questions:
Would Alfred be able to integrate with Agent Zero?
If not, can you add that integration in the future?
Congratulations on an extraordinary project. I'm very impressed and looking forward to future updates.BTW, please do more videos... maybe you could have Alfred create Youtube videos from the blog transcript!
hey thanks so much for the ping! i haven’t worked with Agent Zero yet, but it should integrate yes.
Thank you very much for this breakdown! I’ve been following your posts for a while now because an "Alfred" is something I’ve wanted to build for a long time. I have OpenClaw running on my Raspberry Pi, which is where my agent—also named Alfred—lives. (Coincidence, I swear! I was thinking of Batman while onboarding OpenClaw).
Would it be possible to install your Alfred there as well? Python is already installed, but I’m not quite sure how to connect my OpenClaw instance to your Alfred running on my Pi. That was the one part I didn't quite get from your walkthrough.
yeah in theory is should install on your RPi just fine, just follow the same steps. the pip installer should find openclaw on PATH. if it doesn’t ,let me know. also be mindful that by default Alfred’s Surveyor uses local embeddings (nomic-embed-text) and some ML libraries (HBDSCAN, Leiden algorithm) that may or may not be too much for your RPi. either way let me know how it goes!
Great breakdown
thank you sir!
The layered breakdown makes this easier to grasp.
thank you John!
Thanks so much for sharing this. One thing my ADHD brain is grappling with is where to focus attention among all my subscription channels (Substack, podcast subscriptions, newsletters, articles shared by friends/family, etc.). Have you played at all with using Alfred to filter these channels and make suggestions (i.e curate your own "algorithm")?
I haven't built anything yet, but one idea that we discussed with my wife was to satiate my hunger for knowledge and my curiosity in a way that doesn't find me the screens would be some sort of automated creation feed that curates stuff I would want to read every week. But instead of sending it to my inbox it would automatically send it to a printing service who would then send it to me in some ways so I could get my own weekly magazine fully personalised for my interests all managed by Alfred.
I have no idea if it's possible, but that would be cool
Oh I like that idea to reduce screen time! An aggregate audio channel could also work for me.
I love this idea, I've been building something like this myself with a WordPress App, but I think your Alfred system would be much more streamlined.
That’s EXACTLY how I’ve been thinking of this project.
Douglas Rushkoff points out in his brilliant polemic, Team Human, that normal people are always one communications revolution behind the elite. E.g. priests of antiquity used their literacy to interpret the Bible to an illiterate set of commoners. But when the Guttenberg press was invented and commoners became literate, elite now controlled what books got published. It goes on like that through the telegraph, radio, television, the Internet, and now we’re seeing it beginning for social media, and because things are moving so fast, maybe even for agentic AI as well.
Beginning right here, in this comments section.
As those with the most systemic power now have access to build AI models and influence the world through their weights, we have an ability to build software. This is why we’re hearing about the death of SaaS right?
But it could (must?) also be the death of the dominance of the previous media revolution.
A good intro to how algorithms work might be the Netflix Doc (based on Rushkoff’s work) The Social Dilemma. If you want to dig deeper, Surveillance Capitalism by Shoshana Zuboff is maybe a little outdated now, but an excellent explanation of how the data that we generate by existing gets converted into advertising dollars by companies that can build a predictive profile of us.
And now, I’m seeing all these OpenClaw users naturally inclined to connect as much of their personal data to these tools as possible so that they can make predictions.
Well that’s LITERALLY democratizing the behavioral prediction algorithms and not to get grandiose, but there could be a massive boost in wellness and agency amongst the broader population if tools like Alfred handed that power to normal people.
Meanwhile, the tools themselves “go meta” on the previous communications revolution by bundling feeds and making a single umbrella under which the old world is contained. Just exactly as you suggested.
You had me at the word butler ;)
Love this. I'm actually building something similar, a marketing automation agent for indie devs that handles App Store optimization, Reddit launches, and content scheduling. The hardest part isn't the AI logic, it's defining the guardrails so the agent doesn't go rogue on your behalf.
Great writeup! I'm trying ZO now. You may want to check this out - Obsidian has a new beta headless version to sync from the cli, no need for the desktop version: https://obsidian.md/help/headless
I love how you architected Alfred. It is based on how humans think and do the work without AI, then how humans translate that work with the use of AI tools and any behavior change we, humans must make to truly make AI work for us. I will be attempting to install this! :-)
This is great workflow but Omi presents significant privacy risks as it is uncontrolled recording and storage in cloud. Is there an alternative setup you would recommend?
that statement is incorrect.
omi is a wearable microphone that's fully configurable. the default configuration is what you said, which i disabled.
my configuration is:
-> double button tap turns it on/off
→ default on at home, default off everywhere else
- omi service disabled, all recordings go to my own VPS → transcribed → processed → transcription and audio deleted
that said you dont need ambient recording. I do. sending voice notes does the same job. just record a voice recording on your phone and send it to the inbox folder.
but that assumes more clicks and button taps and a conscious decision to log which for me kills the purpose
thanks for clarifying.
Ok, I did everything you said and it's all running. Do I interact w Alfred through the openclaw telegram interface or do i need to set up something separate?
Awesome! I can definitely use an Alfred in my workflow. Great framework
Very comprehensive, thanks for putting this together.
P.S. One of my favourite words is butler rather due to the historical stories. Such a great analogy. :)
Think of it as an agentic butler: organizing facts, reminders, and tasks across your digital life, reducing mental clutter.
The open-source repo makes it even more exciting, letting anyone experiment and customize their own “butler” for daily life.