Hey HN — we built SlackClaw after spending way too long self-hosting OpenClaw for our team's Slack. Turns out most of the pain had nothing to do with the agent itself.
It was all the stuff around it: rotating Slack tokens, choosing between Socket Mode and HTTP, babysitting a VPS, chasing down why events randomly stop arriving, and trying to keep up with OpenClaw's release pace (they shipped three times in one week recently). Every guide we found got us through the happy path. Nobody talks about month six, when something quietly breaks and you're debugging OAuth scopes on a Friday night.
We kept thinking "someone should just host this properly" and eventually realized we were going to be that someone. So we built a managed OpenClaw-for-Slack layer:
One click to add to your workspace. No Slack app setup, no tokens to wrangle.
Each team gets its own dedicated server (8 vCPU, 16GB) — not shared. Your data stays on your instance.
3,000+ integrations through OAuth. Connect GitHub, Linear, Jira, Notion, Gmail, HubSpot, whatever your team already uses.
"Skills" are just plain-English markdown files describing what you want done. No workflow builder, no drag-and-drop triggers.
Pricing is per action, not per seat. Your whole team uses it, you only pay when it actually does something.
Under the hood it's a full OpenClaw instance with browser, terminal, filesystem, persistent memory, and hybrid search for context recall. Powered by Claude — we never use your workspace data for training.
There's a free tier with $100 in credits if you want to kick the tires. The use cases that surprised us most so far: teams using it for automated standups, incident response playbooks, and onboarding new hires (one command sets up channels, shares docs, sends intros, schedules 1:1s).
Happy to dig into architecture, security, or any of the weird OpenClaw-on-Slack edge cases we ran into building this.
interesting -- so ive been super lazy about it all - as in instead I've just told the agent to make memeory md files -- and do things that way. And it reads through files when it is looking for memories etc). They are like well structured for it etc. Kinda almost like how coding agents read through codebases.
This is marketing, there will be an advert posted in the banner later once the stealth product is ready. Plus also, AI inference costs have plummeted in the last ~6 months - whih is why so many AI services now have such generous free teirs.
AI Writer is a Notion-style editor with AI-powered features. You can use the prompter at the top to fill in the essay, and then use the notion like editor to edit and collaborate with the AI.
Built with javascript libraries, an MMORPG that can hold thousands of players in an open world RPG - inspired by minecraft and stardew valley with elements of WoW. Running on Kubernetes on the backend.