Your catalogue, open to AI shoppers.
See howYour marketing data, in the channel.
Ask a question where the team already is, and get the answer in thread. Anomalies, digests and scheduled reports arrive without anyone asking.
Ask next: Which campaign drove it? • Compare to last month
A step inside Cresva rather than a way to sign up. Not signed in? The link stops at login and brings you back.
Six things that show up on their own
These are scheduled jobs, not notifications you configure one by one. The cadence beside each is the schedule it actually runs on.
Anomalies, as they surface
A metric that moved further than its own history says it should. The message carries the z score, a severity, the brand and the company, and the UTC minute it was detected. Grouped per channel, so one post covers everything found in that window.
The daily spend digest
One company summary, then a line per brand. Each carries a freshness stamp in minutes, so you can tell at a glance whether you are reading this morning's numbers or a stalled pipeline.
Scheduled reports, delivered
When a report finishes, it arrives with a View Report button rather than a link buried in prose. The greeting is time aware, so a report that lands at 09:00 does not open by telling you it is the end of the day.
The weekly safety receipt
A week in one message: days all safe, total spend monitored, and every cap hit or failure worth naming. When there is nothing to report it says so plainly, which is the point of a receipt.
The daily budget guardrail
Meta spend for the day crossing the daily cap set for the company. The message carries the spend, the cap, the overage, and the percentage of cap that overage represents, with the UTC minute it was detected. Meta only, today; nothing posts for a company with no cap configured or no connected Meta account.
The per-brand anomaly rundown
A company-level summary, one section per brand, of the most notable anomalies from roughly the last two days. Repeated content is suppressed within a mute window, so an unchanged summary does not post again on every run.
What you would actually ask
Mention the app in a channel or send it a direct message. It answers in a thread under your message, so a channel can hold several of these at once.
“What was our Meta spend last week?”
Mention the bot in a channel or send it a direct message. It posts a placeholder immediately and replaces that message in place once the answer is ready, so the thread never fills with progress noise.
“Why did ROAS drop on Tuesday?”
The same agent that answers in the dashboard answers here, with the same tools. It reads the data itself rather than searching for a saved report that happens to cover the question.
“Which campaigns are worth more budget?”
Answers arrive threaded under the message you sent, so a channel can hold several conversations at once without any of them turning into a wall of replies.
Where the line is
Answers in Slack are text; charts and uploading a file are dashboard work, and Slack does not pretend otherwise. Most proposed actions you can approve right from the message: see “How it is secured” below for exactly which ones stay dashboard-only.
A workspace, scoped to one company
Every request is verified
Every request Slack sends is checked against an HMAC-SHA256 signature over the raw body, inside a five minute replay window, before any of it is read. A request that fails is refused rather than logged and processed.
Thread replies are rate limited per brand per hour, so a busy channel cannot turn into an unbounded bill.
Tokens and tenancy
The workspace token is encrypted with AES-256-GCM before it is stored, and the column that holds it is classified as sensitive in the schema.
A Slack request can only act inside the company whose connection carries that workspace id. Nothing resolves a brand or a company globally, even when a company has connected more than one workspace.
Every Slack user is a real Cresva user
Every Slack user links their own Cresva account before the app answers anything for them, and that link is re-checked live on every message, not cached at connect time. An unlinked user gets a prompt to connect their account, never a peek at your data. A Slack guest is refused outright, even if they are otherwise a member.
Approving from Slack
You can approve or reject a proposed action right from the Slack message, if your Cresva role allows it there, the same role and self-approval rules the dashboard enforces, checked fresh at the moment you click. Anything irreversible, needing an explicit send confirmation, or above your brand's configured dollar threshold skips the button entirely and links to the dashboard instead. An unlinked Slack user sees a connect prompt where the buttons would be, never a working Approve.
Slack, Claude and ChatGPT reach the same intelligence layer
This page is the Slack door, where the team already works. The others put the same layer inside an assistant. They are ways in rather than separate products.
Questions
Do I need a Cresva account first?
Yes. Connecting Slack is a step you take from inside Cresva, not a way to sign up. The install flow requires an existing login and a company, and it has to finish in the same browser that started it. If you are not signed in, the connect link sends you to the login page first and returns you afterwards.
Which brand does it answer about?
Whichever brand your own Cresva account has access to. If that's one brand, every question is answered about it, no setup needed. If you have access to more than one: a company admin can map a channel to a brand in Slack settings, so questions asked there are answered about that brand; failing that, you can set your own personal default the first time it asks (there's a button to remember your answer); failing that, name the brand in your question. Ask without naming one and none of those apply, and it tells you which brands it could mean rather than guessing.
Can it post charts?
No. Answers in Slack are text. Charts are a dashboard feature.
Can I approve actions from Slack?
Yes, for most proposed actions, if your role on that brand allows approving. The button runs the same approval check the dashboard uses, checked again at the moment you click, so a role that's since changed is honored, not the role you had when the message was posted. Anything irreversible, needing explicit send confirmation, or above your brand's configured dollar threshold has no button at all in Slack: it links to the dashboard instead.
Does it remember the conversation?
Within a Slack thread it carries recent context, and a reply in a thread is read against the message it replies to. It does not share history with the dashboard: a question asked in Slack does not appear in your dashboard threads, and the reverse is also true.
Can one company connect two Slack workspaces?
Yes. Connecting a second workspace adds it rather than replacing the first: each is tracked as its own connection, keyed to that workspace. A message from either workspace is answered using that workspace's own connection; nothing is pooled between them.
What can it see?
The same data the dashboard can see for that company, because it is the same agent behind both. It reads only the workspace it was installed into, and only within the company that installed it.
What does it need from Slack?
Permission to read mentions, post messages, list channels, and look up users. It does not request permission to read your channel history.
Ask in the channel. Answer in the thread.
Takes a minute, and needs an existing Cresva account.