Get the weekly radar
One short email with the freshest products, every week.
Yasmine lives in Slack and finishes real work — drafted emails, checked finances, fixed bugs.
Grounded in available product and source data
Forty unread emails do not answer themselves, but Yasmine, an AI coworker that lives in your Slack, does exactly that — reading every message and drafting a reply for each one directly in Gmail, left waiting rather than sent out automatically.
That same reach-and-finish pattern shows up in finance and code too, not just inbox cleanup: Yasmine goes through a workspace's Stripe account and spreadsheets to report what came in, what went out, and flags a charge that looks wrong, and separately can find a bug, fix it, and show the exact code change before anything ships — nothing goes live until someone approves it.
What keeps that reach in check is a workspace-defined Allow, Ask, or Block policy applied to every tool action: an action marked Ask waits for explicit approval inside Slack, and if that approval never comes, the request is denied rather than assumed fine. Just as specific is what the site says Yasmine doesn't do — it does not train its own models on a workspace's content, does not expose raw API keys or tokens (those stay in encrypted, tenant-scoped storage), does not act without approval where policy requires it, and does not let data cross between separate workspaces.
The site's own comparison table places Yasmine against two named alternatives, Viktor and Claude Tag — Yasmine positioned as a per-channel coworker with its own memory and 2,000+ app integrations available now with no enterprise gate, against Viktor's single org-wide employee model and Claude Tag's beta-only, Team/Enterprise-gated access. Setup itself is short: add Yasmine to Slack through OAuth, then connect whichever tools are already in daily use, from Gmail and Notion to GitHub, Stripe, and Salesforce.
Four plans are available, from a free 7-day trial up to a Dev tier running on dedicated Fargate compute for heavier development workloads, with Pro sitting in the middle as the plan that adds the option to connect a supported API key or account without an added token markup. The page reviewed does not confirm SOC 2 Type II certification as complete — the site states it as on the roadmap rather than finished, worth checking directly before relying on Yasmine for a compliance-sensitive workspace.
Only where the workspace's policy allows it — actions set to Ask must be explicitly approved in Slack before they run, and if nobody responds, the action is denied by default rather than proceeding unattended.
No — the site states this directly: Yasmine does not train its own models on workspace content, and requests are instead processed by the selected model provider under that provider's own terms for the chosen route.
The site's own comparison trades catalog size for structure: Yasmine runs as a separate coworker per Slack channel with its own memory and an explicit Allow/Ask/Block policy per action, rather than one shared assistant working across an entire organization at once.
Yes — disconnecting an integration or pausing a user blocks new access right away, though the site notes a request already in flight at that moment may still finish running.
Not fully confirmed yet — the site lists SOC 2 Type II as on its roadmap rather than something already completed, so that's worth verifying directly before deploying it somewhere compliance is a hard requirement.
Not necessarily — included model routes run on Yasmine's own credits, and only the Pro tier and above add the option to connect a separately supported API key or account on eligible routes.
These 7 comments were collected from the discovery source. ProductVisit does not host replies in Release A.
Really interesting concept .How do you handle long term memory across channels? Can teams choose what Yasmine remembers or forgets?
3 source upvotesI like that it run on the user's own Claude subscription. That feels like a smart way to give people more control over privacy and costs . Nice touch.
3 source upvotesHi Product Hunt, We built Yasmine because we kept watching small teams pay for Claude, love it, and still do all their busywork by hand. A chat window isn't a coworker. What makes Yasmine different: One coworker per channel. Add her to #finance and #marketing and you get two separate coworkers with separate memory and separate context. She remembers last week's decisions instead of starting cold every time. Your Claude, your data. Yasmine runs on the Claude subscription you already pay for. Conversations go to your own Anthropic account. Never pooled, never used to train anything. If your team already pays for Claude, there's nothing extra to buy on the AI side. Actually isolated. Every workspace gets its own dedicated environment for compute, storage, and memory. No shared infrastructure between customers. It's also why she can safely run code on the Dev plan. She asks first. There are 500+ tool integrations, but every consequential action, like sending an email or touching your CRM, needs your approval right in the channel. You set per-tool allow, ask, or block controls. Scheduled tasks. Tell her "every Monday at 9am, post our metrics summary" and she handles it while you sleep. Setup takes about two minutes. Install, connect Claude, say hello. Free 7 day trial, no card required. Use code PH20 for 20% off your first month. We're here all day and happy to answer anything, especially the hard questions about isolation and data handling. We'd also love to hear what you'd hand your coworker first. Show more
1 source upvotesThe per-channel memory idea is genuinely useful, way more flexible than a single bot. Hooked it up to my docs and email in a couple minutes and it actually pulled the right context when I asked for a recap.
Finally tried Yasmine in our team Slack and what stood out was the per-channel memory. It actually remembered context from yesterday's thread without me re-explaining. Genuinely useful.
The idea of each Slack channel becoming its own AI teammate with separate memory is genuinely useful. Setup felt pretty smooth and it pulled the right context without me having to over-explain.
This is cool. Can a Yasmine in one channel query another channel's Yasmine for context?