OpenClaw 3.24 Update: OpenAI-Compatible API, Teams, and Skills UI
Prefer to watch? Watch on YouTube · Subscribe for new builds · Read the transcript
In this video I go through the OpenClaw 3.24 release feature by feature, starting with the one I think matters most: the OpenAI compatibility layer. It's for anyone already running OpenClaw, or anyone who's been thinking about building with AI agents and wants to know what this release changes in practice.
I recorded this in March 2026, when 3.24 was the third major OpenClaw release in three days (3.22 came out on a Monday, 3.23 fixed its bugs that night, and then 3.24 shipped). OpenClaw changes fast, so check the current docs before you rely on any of the details below.
What I cover in this video
- 0:00 Three major OpenClaw releases in three days
- 0:46 The OpenAI compatibility layer and its three endpoints
- 2:17 The
/toolscommand for checking what an agent can do - 4:34 Microsoft Teams integration
- 5:20 One-click skill installs and the redesigned skills UI
- 6:52 Docker container commands, Discord thread naming, and Slack buttons
- 9:10 Workspace file preview, the macOS config redesign, and security fixes
- 10:41 Recap and where to start
The OpenAI compatibility layer
This is the one I care about most. OpenClaw now exposes three endpoints:
/v1/models
/v1/embeddings
/v1/chat/completions
OpenAI's API uses the same three endpoints. That means any app, tool, front end, or automation built to talk to the OpenAI API can talk to your OpenClaw agent instead. That covers RAG tools, custom front ends, and automations that call the OpenAI API.
The difference is that you aren't talking to a generic model on someone else's server. You're talking to your own agent, with its memory, skills, and contacts. I'd been building with OpenClaw for about six weeks when I recorded this, and one of my biggest frustrations was that I could only reach my agent through Telegram, Discord, or the terminal. Now anything that works with OpenAI's API works with OpenClaw.
If you've been wanting to connect a front end to your agent, you don't need a custom API integration anymore. Point the OpenAI SDK in your app at your OpenClaw URL and you're done. That used to take me a weekend. Now it takes about five minutes.
Seeing and managing what your agents can do
The /tools command
We run 13 agents on a single Mac mini, so figuring out which tools and skills each agent has has always been a pain. You'd dig through config files, skill directories, and skill markdown files. Now you type:
/tools
and get a compact list of every tool and skill the agent has. The control UI also has a live section that shows what's available in real time. I've lost hours debugging problems that turned out to be a skill that wasn't loaded, so this one would have saved me a lot of time.
One-click skill installs
OpenClaw's skill system lets you add things like coding agents, GitHub issue trackers, weather tools, and Whisper transcription. Installing them used to be hard for beginners. You had to figure out the dependencies, install them by hand, and sort out API keys yourself. Now bundled skills show you which dependencies are missing and offer to install them in one click.
The redesigned skills section
The skills section of the control UI has been completely redesigned. It now has:
- Status tabs to filter by ready, needs setup, and disabled
- Detail views you can click into, with a requirements checklist
- Toggle switches to turn skills on and off
- API key entry right in the UI
- Source metadata so you know where a skill came from
Someone brand new can now open the control UI, browse skills, install what they need, add their API keys, and get running without opening a terminal.
Workspace file preview and macOS config
In the control UI you can expand file rows and preview your agent's markdown files in the browser, including your soul, memory, and tools files. That lets you see your agent's personality, what it remembers, and what tools it has without opening the terminal. The macOS app's configuration screen also swapped its horizontal pill navigation for a sidebar with collapsible sections. It's a small change, but it makes the app easier to use every day.
Teams, Slack, Discord, and Docker
Microsoft Teams: This release adds full Teams SDK support, with streaming replies, welcome cards, typing indicators, AI labeling, and message editing and deleting. Your agent shows up as a proper Teams bot. I see this as an enterprise play. Teams has over 300 million users, so if you're consulting or selling agent setups to companies, you can put the agent in a tool they already use instead of asking them to install something new.
Slack interactive replies: When your agent's response includes options, Slack now shows them as real buttons and selects instead of plain text.
Discord auto thread naming: Threads used to be named after the first message, which looked messy. Now the AI writes the name, so a question about cron jobs gets a thread called something like "setting up cron jobs."
Docker container commands: A new container flag lets you run OpenClaw commands inside Docker or Podman containers. If you deploy on a DigitalOcean droplet, a Hetzner box, or another cloud provider, managing your instance from inside the container is now much easier.
Security and reliability fixes
- A media sandbox bypass was closed. Before the fix, files could potentially escape the media root directory.
- A Docker fresh-install loop that was tripping up new users was fixed.
- Channel startup is now isolated, so one broken channel doesn't block the others. Before, one misconfigured Discord channel could stop your Telegram and Slack agents from starting too.
Key takeaways
- The OpenAI-compatible endpoints let tools built for the OpenAI API connect to your own OpenClaw agent, memory and skills included.
- The
/toolscommand and the redesigned skills UI make it much easier to see what each agent can do and fix setup problems. - Teams support, Slack buttons, and Discord thread naming make agents work better where people already chat. Teams matters most for business clients.
- Docker container commands and isolated channel startup make server deployments more reliable.
- You don't need to learn every feature. Pick the one that fits what you're building, go deep, and build something with it. If you're new to agents, start with my AI agents overview.
FAQ
What does the OpenAI compatibility layer in OpenClaw 3.24 do?
OpenClaw exposes /v1/models, /v1/embeddings, and /v1/chat/completions, the same endpoints the OpenAI API uses. Apps and tools built for the OpenAI API can point at your OpenClaw URL and talk to your agent, with its memory and skills.
What does the /tools command do in OpenClaw?
Typing /tools shows a compact list of every tool and skill an agent has. That way you don't have to dig through config files and skill directories to find out why something isn't working.
Can OpenClaw agents run in Microsoft Teams?
Yes. Version 3.24 added full Teams SDK support, with streaming replies, welcome cards, typing indicators, AI labeling, and message editing and deleting, so your agent shows up as a proper Teams bot.
Do I need the terminal to install OpenClaw skills after 3.24?
Not for bundled skills. The control UI shows which dependencies are missing and offers to install them in one click, and you can enter API keys and turn skills on or off right in the UI.
Transcript
Read the full transcript
0:00OpenClaw 3.24 just came out and here is everything you need to know about it. I need you to understand something first. This is the third major update in 3 days after being silent for 8 days straight. Monday we got 3.22, Monday night they dropped 3.23 that fixed what 3.22 had bugs in, and then today we just got 3.24. And the OpenClaw is on an absolute tear right now. I don't think you realize how fast things are moving. So if you're building with AI agents or if you're running OpenClaw and you've been thinking about getting into this space, you need to pay attention because this update changes the game for many people that are building with OpenClaw. I'm going to walk you through every single feature that matters. So I'm going to start with
0:46the biggest one right up front. This one had me pacing around my office back and forth figuring out how I was going to implement it, the OpenAI compatibility layer. So I'm going to break down what this means because it is a big one in this latest release. OpenClaw now exposes three endpoints, version one models, version one embeddings, and version one chat completions. If those sound familiar, it's because those are the same endpoints that OpenAI uses, the same ones. So what does that actually mean for you? It means that any app, any tool, any front end, anything that was built to talk to the OpenAI API can now talk to your OpenClaw agent instead. So if you think about that for a second, every rag tool out there, every custom
1:32front end that was built by ChatGPT, every automation that hits the OpenAI API can now point all of that at your OpenClaw instance and it just works. This is not a small thing. This opens the floodgates. The entire OpenAI ecosystem of tools and the apps and the integrations, all of that now can plug into your personal AI agent, your agent that has your memory, your skills, all of your contacts. This isn't some generic model sitting in someone else's server. Your agent now has access to all this. And I've been building with Open Claw for about 6 weeks now, and one of the biggest limitations was always, "Okay, great. I have this incredible agent, but I can only talk to it through Telegram or Discord or the terminal."
2:17Now, I can hook it to this and literally anything. If it speaks OpenAI, it speaks Open Claw. That's it. And I love that. And for builders inside Shipping School who've been asking about connecting their front ends to agents, well, this is your answer. You don't need to build a custom API integration anymore. Just point your app to the Open Claw URL within the OpenAI SDK, and that's it. That used to take a weekend project. Now it's 5 minutes. The next feature, the /tools command, this might sound small, but it's not. If you're running multiple agents like I am, we run 13 agents on a single Mac mini, and you know the pain of trying to figure out what each agent can actually do, what skills it does, and what does it have access to, what tools are available. You'd have to dig through config files, check the skill
3:02directories, read through the skill markdown files, and you don't have to do this anymore. Just type /tools, and you get a clean, compact view of everything your agent has access to. Every tool, every skill, everything laid out and in the control UI. So, now there's a live section that shows you what's available right now in real time. I can't tell you how many hours I've spent debugging issues that turn out to be a skill not being loaded or a tool not being available. The /tool command alone would have saved me a ton of time in the past. So, get this one. It's a really cool update. Real quick, before we keep going, if you're watching this and you want to actually build with some of these tools, not just watch videos about them, then you're going to want to check
3:48out our community down below, Shipping School. We have a full Claude code course, a full Open Cloud course, and four live boot camps every single week where we actually help you get set up from scratch. Like actually set this thing up. Not just watch a tutorial and figure it out by yourself. And we also provide one-on-one coaching so you could book a call with me. We can share screens and I can help you get Cloud Code or Open Cloud running on your machine. That's it. No fluff. I built this community because watching YouTube's only gets you so far. We launched it just 3 days ago and we have over 55 members. You need people around you who are actually building, people who hold you accountable, and coaches who can help you when you get stuck. I'll put the link in the description down below. Get in now before the price
4:34goes up. Number three is Microsoft Teams integration. Now, I know what you're thinking when you hear that, Teams, come on, we're past that. But hear me out. This is enterprise play right here. Full Teams SDK support with streaming replies, welcome cards, typing indicators, AI labeling, the whole thing. You can edit and delete messages. Your agent shows up as a proper Teams bot. Why does this even matter? Because if you're trying to bring AI agents into a company or into a real business environment, Teams is where those people live. Slack gets all the hype in the tech world, but Teams has over 300 million users. And now you can drop an Open Cloud agent right into that environment. Your agent, with your skills, your memory, your contacts, sitting right there in the Teams chat next to your co-workers. So, for anyone
5:20building with AI and you're building a consulting business or for anyone trying to sell agent setups to companies, this just made your pitch to 10 times easier to those businesses. You're not asking them to install some new app or launch some new platform. You're putting the agent they already work, you're putting them to work in that kind of ecosystem environment. Feature number four, and this one is close to my heart, is the one-click skill install. So, Open has this amazing skill system and you can now add coding agents, GitHub issue trackers, weather tools, transcriptions with Whisper, and all kinds of stuff. But, the install process used to be not beginner-friendly, right? You had to know what dependencies were needed, install them manually, figure out API keys, and hope everything works. Now,
6:06the bundled skill shows you exactly what dependencies are missing, and they offer to install them for you in one click. Speaking of the UI skills, feature number five is like the entire skill section in the control UI got overhauled completely. Status tabs, so you can filter by what's ready, what needs setup, and what needs to be disabled. Click to detail logs that show you everything about a skill, requirements checklist, toggle switches to enable and disable skills, API key entry right in the UE, source metadata, so you know where the skill came from. This is a massive quality of life improvement. Open Claw was already powerful, but now it's actually approachable. Someone brand new can open the control UE, go to skills, see what's available, install
6:52what they need, configure their API keys, and be up and running. No terminal required. Wow, that's huge for adoption. The sixth thing is Docker container commands. There's a new container flag that allows you to run Open Claw commands inside Docker or Podman containers. This is huge for anyone running Open Claw on a VPS or on a server in a containerized environment. You can now manage your Open Claw instance from inside the container without jumping through the hoops needed. So, if you're deploying Open Claw on DigitalOcean droplet or a Hetzner box or any cloud provider, this makes your life a bit easier. Container management just became the first-class citizen. The seventh one is Discord auto thread naming. This one is so nice because previously, when your agent created a thread in Discord, the thread
7:38name would just be the first message, which looked like a mess and it wasn't helpful at all. So now the thread name is AI generated. So if someone asked you your agent about maybe setting up a cron job, the thread gets named something more meaningful like setting up cron jobs instead of the first 20 words of the message. Cleaner Discord experience, more organized, meaning less friction when it comes to building. The eighth one is Slack rich interactive replies. Your agent now renders buttons and selects automatically in Slack. So if your agent's response includes options, they show up as actual interactive buttons instead of just text. Better UX, more professional-looking, and it just makes your agent feel like a real Slack app, not just a tap bot. So feature number nine is agent workspace file preview in the control UI. If you're
8:24struggling to keep up with content, well, I'm about to save you about 40 days worth of work. [music] I built something called Content Machine. It's 10 AI agents that run on the Open Claw orchestration, and they handle everything. Scripts, [music] thumbnails, X posts, blogs, outreach, clips, newsletters, all of it. So I went from 1,000 subscribers to 4,000 subscribers on YouTube in 7 days using this exact system. Every single morning I wake up [music] and the content's already done. I spend maybe 15-20 minutes reviewing and approving them, and I move on with my day. It works for any niche, fitness, finance, real estate, marketing, whatever you are building, and it is 100% completely customizable to your
9:10[music] use case. So you get the mission control dashboard, all of the cron jobs, everything I've built over the last 40 days helping me gain more and more people to subscribe and join the community. So you plug in your own thing, and it molds it to you, it learns how you talk, and it writes [music] so it doesn't sound like AI slop. $97 one time, it's not a subscription. I'll put the link down below, and you'll thank me later. You can now expand file rows and see lazy loaded markdown preview right in your browser. So, your soul markdown, your memory markdown, your tools markdown, you can preview all of them without opening the terminal. You can see exactly what your agent's personality looks like, what it remembers, what tools it has right there in the UI. The 10th installment of this
9:55update is Mac OS app just got redesigned configuration screen. They replaced the horizontal pills navigation with a three sidebar that has a collapse section. Way cleaner, way easier to navigate, especially as more config options get added. It's a small thing, but it does make the daily experience much better. Let me also touch on what the security fixes are because they do matter. They closed a media sandbox bypass where files could potentially escape the media root directory. They fixed a Docker fresh install loop that was causing headaches for new users, and they isolated channel startup, so one broken channel doesn't block all the others from starting. The last one, and this one's been a pain in the butt, but you'd have one misconfigured Discord channel, and suddenly your Telegram and Slack
10:41agents wouldn't start either. Come on now. Well, this got fixed. So, I'm going to bring this up all together, three updates in 3 days. The OpenAI compatibility layer opens up Open Claw to basically the entire ecosystem of tools built for OpenAI. The skills UI redesign makes it approachable for beginners. The Teams integration is an enterprise play. The Docker container commands make server deployments much more real, and all the little improvements like /tools command, the Discord thread naming, the Slack buttons, the workspace previews, they really add up to a much more polished experience. And this is what I keep telling people. Open Claw isn't just an AI agent framework anymore. It's becoming the operating system for AI agents. And the speed they're shipping
11:26at right now should prove a point. 3 days, three major releases, each one with features that actually matter to builders. If you're someone who's been sitting on the sidelines, thinking about building with AI agents, this is your sign. The tools are here, the ecosystem's growing, and it's moving at the speed of light. And if you want help getting things set up, and if you want to see how I run my 13 agents on a single Mac Mini, and how we built the content machine, how we use Open Claw for everything from YouTube scripts to GitHub automation to customer support, check out the community down below. We are at 146 members in the first 14 days of launch. And we just need four more members, and then the price goes up. So, get in before the price goes up. We do six live build sessions every single week, helping you take the next step in
12:11your AI journey. And the link is description, so you can check it out. So, whether you join or not, just start building. It doesn't matter to me. I just want you to use these tools to build a business, to make more money, to feed your family, to buy back your time. Don't wait for the perfect moment. The perfect moment doesn't happen. It never exists. Just jump. Take the leap of faith. It's the time. Get going. And here's the thing I want to leave you with. You don't need to understand every single feature I just covered. Pick one. Maybe it's Open AI compatibility layer, because you've got a front end and you want to connect it. Maybe it's the Skills UI, because you've been struggling with installation. Maybe it's Teams, because your day job runs on Microsoft. Just pick one. Go deep and build something with it. That's how you learn. These LLMs and these AI agents, they're not going to slow you down. The updates are coming faster and faster
12:56every day, which is why we do four to six videos every day on this channel. And the people who are building right now, not just watching, not just following along, but actually building, those are the people who are going to be in an amazing position four to six to eight months from now. Please subscribe to the channel, because you don't want to miss the next update. There might be another 3.25 update in an hour. I wouldn't doubt it. So, please subscribe, hit the notification bell, and we'll see you in the next video. Bye-bye.
Ready to start building with AI?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolKeep reading
OpenClaw 4.1 Update: Rate Limit Failover, /tasks, and Cron Allowlists
OpenClaw 4.1 adds rate limit failover that caps retries on one provider before switching to your fallback, like OpenAI. It also adds a /tasks board for background work, fixes the gateway freezing that showed up in 3.31, lets you limit which tools each cron job can use, keeps raw errors out of chat, and makes /model wait until the current run finishes.
OpenClaw 3.22 Update: ClawHub Default, GPT-5.4, and Breaking Changes
OpenClaw 3.22 makes ClawHub the default place to install skills, adds GPT-5.4 (plus mini and nano) as the default OpenAI model, and adds per-agent reasoning settings, /btw, and an SSH sandbox. Before you update, rename old CLAWDBOT_/MOLTBOT_ environment variables to OPENCLAW_.
OpenClaw 5.19 Update: Tool Plugins, Shared Skills, Browser Dialogs
OpenClaw 5.19 is a reliability release. It brings validated tool plugins, global shared skills, browser modal dialog handling, Codex plugin controls from chat, a cleaner Mac settings app and better restart traces. Together these make agents easier to repeat, share and debug.
