Clade
AI COO that runs your team in tools you already use
644 followers
AI COO that runs your team in tools you already use
644 followers
Clade is the AI Digital COO for software teams. It lives in Telegram, Slack and iMessage - learns your tone, knows your team, and runs the operational work (standups, follow-ups, decisions, accountability) while you stay focused on clients and delivery. What's different: every fact Clade learns is transparent - you see the source, the confidence, and can confirm or correct it. No black-box memory. Five trust levels, from observer to operator. Free in Telegram. 1 min setup.
This is the 2nd launch from Clade. View more
Clade
Launched this week
Clade is your AI COO for teams. It lives in Telegram, Slack, and iMessage, so there's no new dashboard to babysit - you message it like a teammate. Clade chases status updates, keeps client and project work moving, drafts the follow-ups, and flags what's slipping before it becomes a fire. The web app is your cockpit when you want the full picture. Built for teams who'd rather ship than manage.
Free
Launch Team / Built With
Clade
Hey Product Hunt 👋 I'm Rahman, Co-Founder & CEO of Clade.
We first launched Clade on Telegram - an AI Digital COO that lives where you work, learns your team and clients, and runs the ops you keep forgetting: standups, follow-ups, the message you meant to send three days ago.
Then our users told us where they actually run their agencies. Not just Telegram. So today we're launching on iMessage, Slack, and a brand-new web cockpit. Same Clade, now everywhere you already work.
It learns your voice and drafts in your tone, catches the follow-ups that slip when you're juggling five deliveries at once, and runs your standups. No starting over every morning.
And nothing's hidden. Every fact Clade learns shows its source, its confidence, and a button to fix it. No black-box memory.
What you get free:
• iMessage, Slack, Telegram + new web cockpit
• Drafts in your voice, runs your standups
• Automations + unlimited memory
• Five trust levels, from watching to operating
Quick question for the founders here: which channel does your team actually live in? That's the one we built for.
I'll be here all day - tell me where this still falls short.
Try it: https://clade.sh 🙏
Clade
@rahman_bazarov_ Rahim here, the other half of Clade 👋
Rahman covered the what.
Let me add the part I lost the most sleep over: the memory.
The hard problem wasn't making Clade do the work. It was making people trust it enough to actually let it. Our first version remembered everything but people checked every output, so it saved them nothing. We almost shipped a fancy verification queue.
So we rebuilt it around trust, not recall.
Every fact carries a source and a confidence, and you can correct it in one tap. That single shift is what made people stop double-checking and start delegating.
Building this across iMessage, Slack, Telegram and web in one product was the hardest thing we've done. If anything feels rough, tell us, we're both here all day.
Clade
@rahimwws Proud to have built this with you. On to the next chapter 🚀
The "no black-box memory — every fact shows source, confidence, and a button to fix it" is the part that would decide it for me, because an ops agent acting on a wrong assumption is worse than no agent. Concretely: if Clade learns something wrong and drafts a follow-up on it before I catch it, does a correction just update the fact, or does it also walk back actions already taken at the "operator" trust level? I'd want to know the blast radius of one bad fact before handing it standups and client follow-ups.
Clade
@hazy0 Good question. A correction updates the fact right away, and we re-check any pending actions that relied on it before they go out. If something already went out based on the wrong fact, we don't auto-walk it back since that could cause more damage - we flag it to you with what was affected so you decide the fix. Blast radius stays visible, never silent.
Flagging it instead of auto-reverting is the right call — silently walking back a client-facing action could do more damage than the original wrong fact. When you surface that flag, does it land in the same channel/thread where the wrong message went out, or in a separate audit view? If a bad follow-up reached a client, I'd want the correction next to that conversation, not somewhere I have to go hunting.
This is interesting. How does it avoid duplicate syncing when the same project runs across Slack and iMessage?
Clade
@dhiraj_patel5 Good question. The trick is that channels aren't separate stores, they're just different doors into one memory.
Slack and iMessage don't each hold their own copy of the project. They both read and write to the same underlying model of your team and work. So a project isn't "synced" between them, it just exists once and every channel points at it.
When the same fact arrives from two places, we dedupe on the entity, not the message. We resolve who and what it's about, check it against what's already there, and either confirm the existing fact or flag a conflict if they disagree. Same input twice doesn't create two records, it just raises confidence on the one that's there.
Clade
@dhiraj_patel5 we tag facts with their source and timestamp, so if Slack says one deadline and iMessage says another, we don't just pick one silently. We flag the conflict and let you resolve it. Keeps trust high even when inputs disagree.
Transparent memory is a smart move. Most AI tools "remember," and you never know what they are actually storing.
Curious though, how do the five trust levels work in practice? Does "observer" watch, and "operator" take action?
Asking because for a COO-type tool, the line between helpful and intrusive is thin. 😅
Congrats on the launch.
Clade
@wasil_abdal Exactly right - observer just watches and logs, operator can take real actions like drafting or sending. There are levels in between too, so you can dial up trust per agent as it proves itself. Totally agree the line matters a lot for a COO-type tool, that's why every action ties back to a visible fact.
Building around trust instead of just memory feels like the right call - especially with sources and confidence visible.
Curious what happens if two channels create slightly different versions of the same fact. Does Clade try to resolve that, or just surface it?
Congrats on the launch!
Clade
@jared_salois We try to resolve it first - matching on entity, not exact wording, so "Q3 redesign" and "the redesign" map to the same thing. If the two versions actually conflict (like different deadlines), we don't guess, we surface it and let you pick. Thanks for the kind words!
Healy
Congrats team, clean execution across three platforms!
Clade
@ecomnazar Thanks Nazar, means a lot coming from you. More to come!
congrats on the launch!
Clade
@a_6 Thank you, appreciate you checking it out!