My team fully abandoned our attempts to make OpenClaw or Claw-like agents work for us.
We invested a sizable chunk of our R&D budget to setting various Claws up to help our software and QA teams. It took about a week after launch for the team to fully sour on the idea of using these systems.
The biggest feedback we collected was that any tasks they assigned to the Claw systems would turn into tangled messes often requiring significant time investment to understand, and mostly ending in the team scraping the code changes over quality concerns.
The team really struggled to find anything these Claw systems could be tasked with that wouldnt end in poor results, So we ended up scraping the idea entirely. I think it will be quite a few more years until these types of systems are appropriate for the enterprise.
We gave it a few months try. My conclusion was that it did interesting things but is designed completely wrong from the ground up which means it's got a lot of moving parts that are part of the code base that can break for all sorts of reasons.
As a learning exercise it was pretty nice. But since then the plugin/connector ecosystem for things like Claude Code/Cowork and ChatGpt/Codex has first come into existence and then matured to the point where they do most of what made OpenClaw interesting before that was the case. But in a way that can be shipped to lots of users. That wasn't true when I started using it. We played with it for a bit but in the end the code base is a mess, random shit breaks every time you update it, and you end up doing very dodgy shit with it that no responsible CIO would want to sign off on.
The current batch of tools from Anthropic and OpenAI is pretty solid but there are still lots of feature gaps, UX issues, security challenges, etc.
I think a lot of these tools will get into the enterprise in exactly the same way that usb sticks, MS Office, smart phones and other consumer tech got into the enterprise: employees will bring them and use them. Some bosses will tell them off and then make an exception for themselves. Because the promise of not having to do monkey work that can now be automated is unbelievably tempting if you need to do lots of that work. Even if this stuff only half delivers on some of the promises. So, my guess is that this could go fairly quickly. I already see a lot of non technical people that are pretty clued in to things like Claude Cowork. There will be some rogue early adoption followed by more enterprise appropriate solutions. That's already happening.
Here in Germany, there are a few German AI companies that work with big conservative German companies and the public sector (e.g. Langdock and Deepset). The type of organization where privacy and data security concerns weigh heavily. These companies can work with OSS models but OpenAI can do proper data residency in Germany as well if you talk to them. Azure, AWS, and others have very acceptable options. And there's a whole ecosystem of companies that are building on top of that.
True Claude and Codex have copied OpenClaws features and that will continue to happen. But with those systems you are getting tied into a proprietary system.
It is the same argument as do you want Linux vs Windows or Oracle vs Postgres.
Over time they are not going to be that different. But with opensource you always have control of your stack. No risk of it ever getting yanked away or it becoming too expensive. The only reason frontier models are affordable is because of opensource.
I have a quite specific use case. I work for a government agency and we run our AI assistant with models from Azure, which works fine! (Although the models aren't always released with EU data residency)
Our most interesting data and documents are classified (VS-NfD), and I don't know of any good certified AI solutions for that yet.
I would love to experiment more with deeper system integration, but we can't do that without frontier VS-NfD models.
You'd pretty much have to self host on premise or in a fully certified environment for classified stuff like that. That will limit your options of course. But some of the OSS models might still be OK for this.
What kind of use cases where you looking to solve. In general coding, marketing and research projects are the sweet spot with these tools. The Claws have commoditized with Claude Cowork, Hermes etc entering the space.
I don't know about other people, but the name "OpenClaw" immediately signals jank and wrecklesness to me. Like, as a consumer, I will pretty much immediately write it off.
I feel the same way. One of my client insists it’s amazing as a marketing tool, helps go through notifications, draft personalized emails..etc., He asks for user metrics from Telegram and gets report…etc., But I am very conflicted. How do you trust the numbers it returns, how will you catch issues in emails if each one is customized and there is no draft review? It’s not my domain of expertise. So I don’t say anything. I just rationalise it as startup culture.
That's why we built https://lobu.ai which is a multi-tenant implementation for proactive agents like OpenClaw. It has entity based memory for building the org context layer and every channel/user gets its isolated container.
I've never run it, but isn't OpenClaw kind of useless if you sandbox it? I thought the entire point is that it has access to do everything. The danger with OpenClaw IMO isn't so much that your local machine gets hacked (although it's certainly a real danger), the danger is giving sensitive data to something horribly unreliable that can leak it or take actions on your behalf that are very dumb. I can understand (although very much disagree) with individuals running it, but trying to do it in the "enterprise" feels like playing with fire in a bad way.
With agentjail (github.com/LuD1161/agentjail), I've tried to contain coding agents in os-native sandboxes (sbpl for macos and similarly for linux) + policy guardrails evaluated by Open Policy Agent (OPA), policies written in rego.
Protocol aware network proxy coming soon
Then you can match a DSL and block particular network requests.
This ensures you no longer fear --dangerously-skip-permissions and stop babysitting agents
What else would you want to see in this project? Please star the repo, if you like the idea :)
Anything made in a couple of days or even hours is abandoned easily. This is how I felt when I saw the onslaught of OpenClaw tutorials months ago. I wonder how many of those influencers who felt they were so far ahead of the game are still using it and if they still feel the GAME has been SHOCKINGLY CHANGED.
I was a basic user who happened to just TRY OpenClaw. Thankfully it was brief and thankfully I ended my session after being underwhelmed because the very next day, they were on the news for the wrong reasons. There is no way on earth I am trusting this or anything even remotely similar in an enterprise setting. In a way, it was course correction for me - I was super optimistic about AI until I saw how ugly it could get. Nope. Nopedy nopedy nope.
I tried Claude, it was half decent. The real issue is these are pretty much ways for these AI companies to burn your tokens faster. Claude is notorious for this and I honestly can't get any real work done without ensuring it isn't randomly erroring out due to token limits. I don't use Codex so can't really comment. What really worked well for me is - using Codex/any other coding tool to write your own API integrations to work with (Eg. Google Calendar) and run a local server of your choice with MCP in it and use openrouter + some cheap model to get the actual work done. This is overblown, but only because the APIs I consume are non-standard.
I wouldn't want OpenClaw anywhere near a business I was running, even if I was all aboard the LLM hype train. That codebase is a nightmare. Just take a look at the hundred-plus pages of bugs on the Github repo.
My team fully abandoned our attempts to make OpenClaw or Claw-like agents work for us. We invested a sizable chunk of our R&D budget to setting various Claws up to help our software and QA teams. It took about a week after launch for the team to fully sour on the idea of using these systems.
The biggest feedback we collected was that any tasks they assigned to the Claw systems would turn into tangled messes often requiring significant time investment to understand, and mostly ending in the team scraping the code changes over quality concerns. The team really struggled to find anything these Claw systems could be tasked with that wouldnt end in poor results, So we ended up scraping the idea entirely. I think it will be quite a few more years until these types of systems are appropriate for the enterprise.
We gave it a few months try. My conclusion was that it did interesting things but is designed completely wrong from the ground up which means it's got a lot of moving parts that are part of the code base that can break for all sorts of reasons.
As a learning exercise it was pretty nice. But since then the plugin/connector ecosystem for things like Claude Code/Cowork and ChatGpt/Codex has first come into existence and then matured to the point where they do most of what made OpenClaw interesting before that was the case. But in a way that can be shipped to lots of users. That wasn't true when I started using it. We played with it for a bit but in the end the code base is a mess, random shit breaks every time you update it, and you end up doing very dodgy shit with it that no responsible CIO would want to sign off on.
The current batch of tools from Anthropic and OpenAI is pretty solid but there are still lots of feature gaps, UX issues, security challenges, etc.
I think a lot of these tools will get into the enterprise in exactly the same way that usb sticks, MS Office, smart phones and other consumer tech got into the enterprise: employees will bring them and use them. Some bosses will tell them off and then make an exception for themselves. Because the promise of not having to do monkey work that can now be automated is unbelievably tempting if you need to do lots of that work. Even if this stuff only half delivers on some of the promises. So, my guess is that this could go fairly quickly. I already see a lot of non technical people that are pretty clued in to things like Claude Cowork. There will be some rogue early adoption followed by more enterprise appropriate solutions. That's already happening.
Here in Germany, there are a few German AI companies that work with big conservative German companies and the public sector (e.g. Langdock and Deepset). The type of organization where privacy and data security concerns weigh heavily. These companies can work with OSS models but OpenAI can do proper data residency in Germany as well if you talk to them. Azure, AWS, and others have very acceptable options. And there's a whole ecosystem of companies that are building on top of that.
True Claude and Codex have copied OpenClaws features and that will continue to happen. But with those systems you are getting tied into a proprietary system. It is the same argument as do you want Linux vs Windows or Oracle vs Postgres. Over time they are not going to be that different. But with opensource you always have control of your stack. No risk of it ever getting yanked away or it becoming too expensive. The only reason frontier models are affordable is because of opensource.
I have a quite specific use case. I work for a government agency and we run our AI assistant with models from Azure, which works fine! (Although the models aren't always released with EU data residency)
Our most interesting data and documents are classified (VS-NfD), and I don't know of any good certified AI solutions for that yet. I would love to experiment more with deeper system integration, but we can't do that without frontier VS-NfD models.
You'd pretty much have to self host on premise or in a fully certified environment for classified stuff like that. That will limit your options of course. But some of the OSS models might still be OK for this.
What kind of use cases where you looking to solve. In general coding, marketing and research projects are the sweet spot with these tools. The Claws have commoditized with Claude Cowork, Hermes etc entering the space.
what did you guys settle on? did you abandon the idea of personal agents completely or did you move to another harness and/or approach?
I don't know about other people, but the name "OpenClaw" immediately signals jank and wrecklesness to me. Like, as a consumer, I will pretty much immediately write it off.
I feel the same way. One of my client insists it’s amazing as a marketing tool, helps go through notifications, draft personalized emails..etc., He asks for user metrics from Telegram and gets report…etc., But I am very conflicted. How do you trust the numbers it returns, how will you catch issues in emails if each one is customized and there is no draft review? It’s not my domain of expertise. So I don’t say anything. I just rationalise it as startup culture.
We also tried to use OpenClaw for teams but unfortunately OpenClaw gateway is not designed as multi-tenant. Peter also confirms it: https://x.com/steipete/status/2026820269050855757
That's why we built https://lobu.ai which is a multi-tenant implementation for proactive agents like OpenClaw. It has entity based memory for building the org context layer and every channel/user gets its isolated container.
Interesting. A different approach. I give every team mate their own openclaw on their VM. Would love to chat and compare notes.
are you based on openclaw? Seems similar to https://flueframework.com/
We use Pi harness, same agent that powers Flue
I've never run it, but isn't OpenClaw kind of useless if you sandbox it? I thought the entire point is that it has access to do everything. The danger with OpenClaw IMO isn't so much that your local machine gets hacked (although it's certainly a real danger), the danger is giving sensitive data to something horribly unreliable that can leak it or take actions on your behalf that are very dumb. I can understand (although very much disagree) with individuals running it, but trying to do it in the "enterprise" feels like playing with fire in a bad way.
I think it depends on the use-case. From what I have seen, seems to be used most for coding, research and marketing.
With agentjail (github.com/LuD1161/agentjail), I've tried to contain coding agents in os-native sandboxes (sbpl for macos and similarly for linux) + policy guardrails evaluated by Open Policy Agent (OPA), policies written in rego.
Protocol aware network proxy coming soon Then you can match a DSL and block particular network requests.
This ensures you no longer fear --dangerously-skip-permissions and stop babysitting agents
What else would you want to see in this project? Please star the repo, if you like the idea :)
Firecracker microVMs are safest. That's what I added to my repo; see above.
Truthfully, this sounds like you're showing folks the magic code to get inside various arcade claw machine games.
I dont understand.
Anything made in a couple of days or even hours is abandoned easily. This is how I felt when I saw the onslaught of OpenClaw tutorials months ago. I wonder how many of those influencers who felt they were so far ahead of the game are still using it and if they still feel the GAME has been SHOCKINGLY CHANGED.
This is true, the market has commoditized.
I was a basic user who happened to just TRY OpenClaw. Thankfully it was brief and thankfully I ended my session after being underwhelmed because the very next day, they were on the news for the wrong reasons. There is no way on earth I am trusting this or anything even remotely similar in an enterprise setting. In a way, it was course correction for me - I was super optimistic about AI until I saw how ugly it could get. Nope. Nopedy nopedy nope.
Do you use Claude or Codex, how different is it really?
I tried Claude, it was half decent. The real issue is these are pretty much ways for these AI companies to burn your tokens faster. Claude is notorious for this and I honestly can't get any real work done without ensuring it isn't randomly erroring out due to token limits. I don't use Codex so can't really comment. What really worked well for me is - using Codex/any other coding tool to write your own API integrations to work with (Eg. Google Calendar) and run a local server of your choice with MCP in it and use openrouter + some cheap model to get the actual work done. This is overblown, but only because the APIs I consume are non-standard.
I wouldn't want OpenClaw anywhere near a business I was running, even if I was all aboard the LLM hype train. That codebase is a nightmare. Just take a look at the hundred-plus pages of bugs on the Github repo.
What agentic tools do you trust? Claude, Codex, Cursor, Hermes.
With all the openAI token spend they get, where do you think the project will be in a year?
OpenClaw needs to be built from the ground up.
I created Atom based on current research on agentic systems and am looking for maintainers & testers to help out.
https://github.com/rush86999/atom
https://github.com/rush86999/atom/blob/main/docs/features/at...
Why did you name it Atom?
So many ai & coding related products is named Atom
Autonomous Task Orchestration Management Agent === ATOM Agent
There are already many different rewrites of openclaw.