Michael O'Nan

After you have the workflow defined ask your agent to extract parts of the workflow that can be handled in code to avoid hallucinations. I find it’s better to have it execute scripts then fully taking on every action. Also might consider a specialized agent for parser writing that is seperate and called by the main agent receiving slack messages.

I haven't used OpenClaw yet but I have thought of a use case and I'm looking for feedback I need • Always on agent waiting for new msgs in Slack • GitHub integration • A bit of custom integration work with my app • Run on a cloud server instead of Mac Mini I want it to monitor a Slack channel where my app sends failed @HandHistory_com poker history import errors Then the agent would see that error, download the poker history file, write a new parser that works for it using Claude Code, push the changes to GitHub (where it will auto-deploy thanks to Laravel Forge) Next it will log in as an admin and reimport the hands that were uploaded (might need to make some API things here so it doesn't need to webscrape or macro this part) Then it will post back to Slack saying it fixed it with a link to the user profile for proof

@steipete @coyington Never experienced this issue with Claude sonnet or opus. Switched to gpt 5.4 and it does it does it frequently. I’m updating the user and soul files to see if that changes anything for me.

@coyington Yeah Anthropic cut us off so no more lying Opus :)

it took a hammer, but i managed to beat emotions into GPT.

@J_dev363 @TheCraigHewitt I use the max plan in Openclaw. x.com/calebhodges/st…

Same here, right now I keeping stuff simple and letting the bot to figure out stuff for itself. Also, most of the issues this I’m seeing is a model choice problem. Kimi is ok but a lot of hand holding, codex is better but it is not as good as opus. If you want true agentic you have to go with opus. But then you cannot use the max plan

my current reality with OpenClaw: I want to use it more I know it's the future But it's so less productive than just using Claude Code and Codex. Doesn't mean I'm not using it. And more importantly, I'm trying to build things with it. Make it more resilient Make it more of a real business tool But it's pushing a boulder up the hill. Those thinking that you just install it and have a 24/7 always on agent doing tons of shit for you are misleading you. It's a ton of work, it breaks a lot, it forget all sorts of shit. But it's the future. We're early, its the right time to put in the reps.

@thekitze Tell openclaw to use a Linux cron instead to call the agent. Problem solved.

you cannot rely on openclaw for productivity yet. maybe in a month or two.

@the_smart_ape Threads in Discord are also great for projects or workflows

if you're not using your OpenClaw agent on Discord yet, do it now. i was running everything through TG. it works, but when you send a task you have to wait for the agent to finish, then send the next one. everything is sequential. one task at a time. Discord has channels. each channel can become an independent agent session. same agent, same workspace, same files, but separate conversations running in parallel. the productivity difference is big. i went from working on one thing at a time to 5-6 projects in parallel. here's how to set it up (takes 10 minutes): 1. go to Discord dev application → New Application → give it a name 2. go to Bot tab → enable Message Content Intent and Server Members Intent under Privileged Gateway Intents → save 3. go to OAuth2 → URL Generator → check "bot" in scopes → in Bot Permissions check: View Channels, Send Messages, Read Message History, Embed Links, Attach Files, Add Reactions → copy the generated URL 4. open that URL in your browser → pick your server → authorize 5. in your OpenClaw config, add Discord: 6. restart gateway. bot goes online. done. I recommend to add a system prompt per channel so the agent automatically knows what project it's working on. no need to re-explain context every time. each channel becomes a specialized agent that remembers everything you discussed in that channel. it reads the same workspace files, has the same tools, same SSH access, but stays focused on its topic. Telegram for quick back-and-forth with your agent. Discord for parallel workflows. use both.

@philkellr I’m not sure why you’re having so many issues but I setup Discord and it works great. The key is telling it what you want and not necessarily how to get there and it surprises me often.

12h in working with OpenClaw All I got so far is frustration It's full of bugs: Telegram integration doesn't work 🤷‍♂️ The web is full of users who claim it's broken without any workarounds WhatsApp sends validation messages to my friends Discord I tried to get working for 2h, both ChatGPT and Claude gave up - I finally got it working by updating OpenClaw 🤯 Steep learning curve All I got so far are python scripts which I built with Cursor and now I painfully need to learn how to somehow connect to each other. I would be much faster writing individual scripts, all writing/reading from a SQLite and then I'd trigger them either with cron, starting manually or – heck even connecting to Discord or Slack I can just code myself much quicker than to debug with ChatGPT or Claude who get the cli options wrong 50% of the time because the project moves so fast. I would spare myself learning a whole ecosystem which seems so buggy that it's probably gonna break with the next update?

@ryancarson I built and open sourced @_ClawController Using it for multiple use cases orchestrating teams of Openclaw agents. Open to feedback x.com/michael_onan/s…

This kind of orchestration is what everyone is trying to build right now. None of us want 1 agent now. We want 1 agent who runs teams of agents. Be interesting to see who solves and ships something with truly delightful UX. Will also be interesting to see if it’s built on top of @openclaw or something else. I don’t think it’s going to be Codex or Claude Code either - it can’t be a one-lab solution. The solution will be a clever mix of closed / open source models + deterministic orchestration.

@ycombinator @usebits_inc @rob0the0nerd Setup is step one. Step two? Managing multiple agents without losing your mind. I built ClawController - an open-source orchestration layer where your lead agent reviews work from other agents and rejects bad output. AI holding AI accountable. x.com/Michael_ONan/s…

OpenClaw is a new viral and scarily effective virtual assistant. But it's so hard to set up that even most engineers give up. @usebits_inc will set up your secure OpenClaw instance on the cloud, batteries included, in 5 minutes. Congrats on the launch @rob0the0nerd and @bailey_wickham! ycombinator.com/launches/POK-k…

@Shpigford @openclaw You might try @_ClawController as it keeps completed tasks in a database so not having as many hook issues. Recurring tasks built in with cron.

i've encounter two core-feature bugs with @openclaw in the last 24 hours that completely kill major functionality: 1. cron jobs don't run 2. broken "hooks" system that prevents auto saving conversations to memory cron jobs were fixed in today's release. broken hooks seems to have a PR ready, but at best will likely be 2-4 days before it makes it in. the broken hooks one is especially brutal as it means my dude basically keeps no memory of any conversations without explicitly telling him to remember it. ☹️

@leonardo_forato @pbteja1998 Use any model with @_ClawController

@pbteja1998 I use Claude Code with GLM 4.7 (not a claude pro subscription). In this case the missionControl will work?

I am blown away by this 🤯 So many people wants early access to MissionControlHQ.ai Talking to each and every one of them to make sure to get it right for their usecases. I have people from all industries - marketers, writers, developers, cyber security experts, teachers - who have a usecase for something like this. It's like a Squad of AI agents working and collaborating with each other to achieve a single mission that you set!! And the best part is... you get to see how they are doing it...

Easily create and manage a team of OpenClaw agents with @_ClawController

Introducing @_ClawController to orchestrate OpenClaw agents.

I built this @openclaw agency management pipeline with @_ClawController, which helps agencies develop content, manage relationships, and manage ad spending.

I built this @openclaw SaaS management pipeline with @_ClawController, which helps SaaS companies with product management, development, branding, SEO, lead gen, and support.

I built this @openclaw trading dashboard with @_ClawController, which helps you be more systematic in risk management, strategy, execution, and market research.

@Shpigford I have had zero success pointing Openclaw to any article. Grok also seems to not be able to access an article as it’s embedded in the post.

@rchase I launched a school this year and used a domain warmup service which drastically increased deliverability. This is the one I used. There’s many others as well. instantly.ai/email-warmup

I bought a new domain handhistory-com and even though it's been registered since 2013 and I've configured SPF DMARC DKIM every email I send goes to Spam folder Any ideas? I don't remember this happening before even with brand new domains

@gowithquo Cool rebrand but the citric yellow is an awful color. grok.com/share/bGVnYWN5…

Our new logo couldn't be simpler. It's our name! We explored hundreds of abstract marks meant to convey feelings about communication, collaboration, and AI. But in a world where so many logos look the same (there are a limited number of shapes out there after all), we realized our greatest asset was right in front of us: Quo. Our new name is short, simple, symmetrical, and clearly us. It stands on its own. Alongside our newly designed wordmark is our new color, citric yellow. It's bright and warm, but also energizing and distinctive, helping our name stand out even more. This is the next chapter of our identity and you'll see it applied across our website, apps, and social media pages.

@Micro2Macr0 @MRBARRIENTOS @grok 3 tasks are available on the $30/month plan as well.

I just found out that they can automate tasks for me so I might be able to literally have it pull up all the news related events and summarize them at the beginning of the day which is something I spend hours doing. I'm thinking I might even be able to have it do automated posting

Officially upgraded to @grok HEAVY! I'll let you know what I think and if it's worth it.