PromptQL - Multiplayer AI that replaces Slack

I'm Tanmai, CEO/cofounder at PromptQL and excited to share the world's first AI native workspace with you. Thank you @kevin for hunting us today!

The team behind PromptQL is Hasura, and we ended up building PromptQL entirely by accident.

Here's the story:

• Be us

• Try to build AI that's accurate with data & tools

• Realize that good context is the bottleneck as AI gets better

• Realize that no one person in the team has all the required context

• Build multiplayer AI (shared threads) so that people can work together in a shared AI thread and automatically capture context

• Move our 70 people team out from Slack to "test the multiplayer experience" on desktop & mobile in Feb 2026

• Things get f***ing weird

• Get team-wide AI psychosis by having a single shared AI connected to everyone and everything

• Realize that we built a Slack for the AI era

So now we don't do DMs or meetings, because the AI knows.

When I ask: "What's alex working on?", its able to look at Alex's work in the workspace and across other tools like slack or github (that is visible to me ofc) and gives me a sense of what's happening, what he's blocked on and how I can help him. I don't disturb Alex and I get an accurate status report. No meetings needed.

Or going further, I can do things that I don't know how to do myself:

When I ask "Help me build this integration for a customer", it uses skills and context from other engineers who've built and reviewed integrations before. Even if we don't explicitly maintain a shared skills repo, I'm suddenly as powerful as my entire team, not just my individual ability.

PromptQL is the first agentic workspace because the whole app is an AI:

PromptQL builds task-specific agents on the fly as users talk to each other. It builds connectors to databases, SaaS tools and internal APIs and learns skills on the go.

Other AI Slack clones (including Slack itself) allow humans and agents to talk to each other. But these agents have to be specifically built and added to the workspace, which we think is just a minor milestone in the path to AGI where AI builds the agents for you with the right security and context guardrails!

We think that tools like Slack are holding us back from doing the most important work that we need to do as teams. We need to be building shared context that we can then apply to the hundreds of tasks in parallel with AI. We need to work together in a way where humans are responsible for context and agents are responsible for execution.

Exclusively for the ProductHunt community, we're giving $1000 in tokens if you want to try PromptQL along with your team!

Please comment here if you're interested and tell us about your team. We'll set up a quick custom 15 minute onboarding for your team and unlock those credits for you that work across all our available models from Fable to Sol to Kimi-K3.

You can also sign up yourself and unlock free credits to play around with it yourself.

PromptQL has been an amazing addition to our AI stack. It's one place that now stores all the context of our team - across various workflows - research, todo lists, marketing, finance, decisions made during workflow. Our slack usage is decreasing daily (we still use it for notifying and general chit-chat) as work has moved to promptql. What I truly love is the cross-model functionality - where I can try the latest models almost the day they are launched and see how they perform on our workflows.

A few feature requests:
- A dedicated section for recurring workflows
- Easier admin functionalities including tracking usage, spends, privileges, etc
- Ability to edit tables and artifacts directly in the canvas
- External teams connect (similar to Slack connect)

Keep on with the great work!

I work at PromptQL, so obviously biased, but here's how I actually use it day to day.

Anyone who lives in Slack knows the FOMO: interesting discussions happening in ten channels at once, and you can't be everywhere. Curing that yourself normally means dedicated engineering time to build something and keep it running. Here it's just... the default. Whatever's interesting or relevant finds you, full context always at your fingertips. You just start a thread and ask PromptQL to find things that might be interesting (define THE interesting ofc) and it just silently works and keeps you updated.

Works outside work too. I use it to trace how a news story actually unfolded instead of getting the echo-chamber version from my feed.

But my favourite thing is how a product gets built now. A thread and the learning becomes the RFC itself. You, your teammates, stakeholders and the AI hash out everything in one place: UX, backend constraints, cost, the actual implementation. One thread, full lifecycle of a feature. I've had the pain firsthand of answering the same questions over and over, explaining why some decision was made months ago, digging through old threads to reconstruct context. Now the feature itself can be Q&A'ed with AI because it watched the whole thing happen. What's more pleasing to a feature builder than not fielding every how/why question, and only getting pulled in when it actually matters?

Even the boring part of shipping (watching adoption, catching issues) gets easier. Hook up PostHog and Grafana and it quietly takes on the operator role. You just approve when needed.

What is nuts about working with PromptQL is how easy is to say:

“Is this thing a problem? Show me data”

“Right, what should we do to fix it”

“Yeah that looks like a good fix, let’s iterate a bit and then deploy it”

“Monitor the problem data every day and let’s check we fixed the problem”

When “doing things” is cheap, making sure you’re not doing useless things is key, and that’s so easy for me now.

interesting idea. what made you take on slack?

biggest strength here is the cross model switching. biggest risk is lock-in once the wiki gets deep.

Congrats on the launch! I like the core bet: humans own context, agents do the work.

But Slack is a muscle memory at this point. How do you see that playing out? Is full migration the goal, or is the real win that team context becomes queryable and agents act on it, and chat just happens to be where it lives? Let's say that my team’s stuff lives in Notion, Google Drive, Jira, and Slack. What does week one look like? Do we move in fully, or does PromptQL read from those tools while we ease over?

the correction/lock mechanism answer covers factual mistakes well, but what about genuine disagreement that isn't a mistake at all - two senior people with different but equally valid judgment calls on how something should work. does the shared thread surface that as visible tension (both views, both names attached) or does it end up flattening into whichever answer got approved first, making the other person's actual position invisible to everyone who reads it later?

I never thought we could replace Slack completely, but when we did and started doing almost all work in PromptQL I realized the true power that a "native AI" experience can bring in an enterprise.

My usual flow when I am stuck is ask PromptQL, and it just knows - because someone else had figured it out earlier and PromptQL had learned it. Everything from "who has office keys" to "how can I access to the latest open-weight model in my local dev" - it JUST knows. Apart from asking questions, I ask it to do things, whenever I find a small product issue - I don't just raise the issue with the engineering team, I ask PromptQL to fix -> show me preview -> inform appropriate team/people for an approval -> deploy.