Hey everyone!
We re excited to share that Agentplace 2.0 launches tomorrow, November 20!
Agentplace 2.0, a platform for building AI-native websites. You can build everything from AI product advisors and consultant sites to AI receptionists, brand agents, and personal AI replicas.
You ll be able to test the live demo for free, and if you d like to dive deeper and explore the full platform, you ll also get access to a promo code with 100 credits to try Agentplace more extensively.
Congrats on the shift to Agentplace, Vlad! As someone with 30 years in IT and Logistics, I’ve learned that the 'perfect agent' is indeed a myth—the real value is in the recovery time when things break.
I really like your 'Edit mode' vs 'Work mode' distinction. In my current project, OmniWatchGuard, I'm dealing with similar challenges regarding DOM-structure stability. When the 'environment' (the website) changes, the agent usually fails silently.
How does Agentplace handle 'environment drift'? If a browser-based agent encounters a UI change in 'Work mode', does it alert the human to enter 'Edit mode' immediately, or is there a self-healing layer?
Great vision on the shared workspace—execution is where the real bottleneck is today!
Agentplace
@omniwatchguard Thank you so much, this really resonates. We think about it very similarly: the goal isn’t a mythical perfect agent, it’s making failure visible and recovery fast. Right now the big priority for us is helping people catch issues quickly, understand what broke, and move back into Edit mode without a lot of friction. A stronger self-healing layer is definitely part of the longer-term direction too.
@polina_semina Glad to hear we're on the same page, Vlad! In logistics, a 'silent failure' is a nightmare—it's much better to have a system that says 'I'm stuck, help me' than one that pretends everything is fine.
That 'low-friction' move back to Edit mode is exactly what will separate the tools people actually use from the ones they just try once. Looking forward to seeing that self-healing layer evolve.
I’ll be keeping a close eye on Agentplace. Cheers to building more resilient automation!
@polina_semina @omniwatchguard Really appreciate and completely agree. A clear “I’m stuck” is always better than a silent failure. That balance between resilience, visibility, and fast recovery is a big part of what we care about, so it means a lot to hear this resonates.
@polina_semina @kaysinb Spot on, Boris! Resilience isn't about never breaking—it's about how gracefully you handle the break.
In 30 years of IT, I've seen too many systems try to be 'perfect' and fail catastrophically. A system that 'knows its limits' and communicates them clearly is a sign of architectural maturity.
Great to see we're building with the same philosophy. Wishing you and the Agentplace team a massive launch week!
Jinna.ai
Congrats on the launch! Agents instruct agents… Does you tool work over a codebase to tailor an agent for it? Asking because generic agent instructions is something I believe Claude itself can generate, wondering how it works in your product
@nikitaeverywhere Great question! Yes, our Builder agent works directly with the agent's codebase. It reads files, edits code, runs commands, checks logs, even takes screenshots of the running preview to verify things look right. So it's not just generating a prompt and hoping for the best. It's iteratively building and refining a full working app with UI, tools, server logic, the whole thing. Think of it more like an AI developer pair-programming your agent into existence, not a prompt generator.
Agentplace
@nikitaeverywhere Thanks! Not just generic instructions, the idea is to shape agents around real workflows, tools, and context. So yes, you can tailor them much more specifically than a one-off prompt generated by a model.
Agentplace
@nikitaeverywhere the magic is in the SKILLS set
Congrats on the launch! What happens when a model gets updated or replaced? How much work is it to re-test and adjust an existing agent?
@ermakovich_sergey Thanks! Good question. On our side, we have internal benchmarks for the Builder agent, so when a new model drops we can test and adapt pretty quickly, usually a day or two. As for the agents users have already built, we don't remove access to older models, so everything keeps working as before. If a user wants to switch to a newer model, we'd recommend testing it on their end to make sure things behave as expected. But nothing breaks automatically.
Agentplace
@ermakovich_sergey, adding to Boris's comment, you can connect any eval tool to enable a controlled change.
Agentplace
@ermakovich_sergey Thank you for support us!
Huddle01 Cloud
Interesting that this is built around workflows and not just 'we have AI now.' Quick question: how easy is it for an ops person to pick this up without looping in engineering every time? That's usually where these things break down.
Agentplace
@shalini_umrao Thank you for bringing this up. That's actually who we built this for. The whole point is that an ops person can build and update agents on their own. There's no code to write, you describe what you want the agent to do in plain text, test it in the same window, and hit publish. If something needs fixing you just open the editor, change the prompt, test, publish again. No pull requests, no deploys, no waiting for engineering
Agentplace
@shalini_umrao Exactly. If every small change needs engineering, adoption usually stalls.
Can Agentplace plug into existing pipelines without restructuring, or do teams inevitably have to adapt their stack?
Agentplace
@athsara Agentplace can plug in natively, so no need to adapt. That being said, AI is a different animal, and we found that to use it to its full power, the process should be different and in many cases simplified.
Agentplace
@athsara Each agent is also an MCP server so you can just make a small one that does one thing and plug it into what you already have. Or if MCP doesn't fit you can add custom API endpoints since it's a full Node.js app under the hood. Either way you don't need to rebuild anything
FinCheck by Trezy
Congrats on the launch! Pivoting away from something that was already working takes guts! respect for that.
Curious: what's the wildest agent someone's built since you made the shift?
Agentplace
@jean_bonnenfant2 Thank you! I know I know I'm biased but my favorite one was a Christmas agent with ElevenLabs voice that we made in team internally. You pick a character like Grumpy Santa or Hip Hop Elf, it talks to you in real time, roasts you a little, and makes a personalized postcard you can send to friends :)
For more serious stuff, there's a competitor researcher agent that goes out, gathers intel on your competitors and comes back with a positioning brief ready to use
Agentplace
@jean_bonnenfant2 @evgeny_sorokin Grumpy Santa was genuinely impressive btw 😄
@jean_bonnenfant_tinct means a lot! Honestly, some of the most fun ones are the weird playful agents people build first, and then a lot of them end up turning into something surprisingly useful.
How do you handle reliability and trust when agents become more autonomous like specifically, what mechanisms exist for debugging, auditing decisions, and preventing silent failures in production workflows
Agentplace
@lak7 Honestly this is an area we're still working on. Right now each agent runs in an isolated VM and you have SSH access from the builder so you can check what's happening. For published agents we don't have built-in logging or audit yet. Since you have full access to the source code though, it's pretty easy to plug in something like Langfuse or Sentry or whatever monitoring you already use. So right now this is more on the user side, but built-in observability is on our list.
@evgeny_sorokin Gotcha! I am using agentplace but it would be cool if there is an docs for its usage. Specifically to edit agent
Agentplace
@evgeny_sorokin @lak7 Awesome! Thank you for trying Agentplace!
Agentplace
@evgeny_sorokin @lak7 we definitely need to make it happen
@lak7 That’s exactly why transparency matters so much to us. If agents are doing real work, people need a clear way to inspect actions, audit decisions, and catch failures early.