Sazabi in Slack: investigate incidents where your team talks

Last verified on Nov 8, 2026.

Most teams already talk about production in Slack. An issue that lives only in a dashboard has to be carried into that conversation by hand, and the answer to "is this real?" ends up in a thread nobody links back.

Sazabi is AI-native observability and security. Its Slack integration puts issues and the agent in the channel itself. People act on an issue where they read it, and ask follow-up questions without opening another tab.

What Sazabi does with Slack

Slack is a chat integration and a shared notification channel. It does three things.

Issue cards. Sazabi posts an issue as a card with inline Resolve, Ignore, Mute and Re-open buttons. Two follow-up events are shaped for Slack threads. Issue updated arrives as a reply under the issue's alert, at most once per issue a day unless severity rises again. Issue reopened arrives as a new message, with what came back threaded under it.

Routing. Which channel gets which card is set by delivery rules on the project's notification settings. A rule names an event type such as Issue triggered and a destination channel, and can add a component, owning-team or severity condition. A rule only adds delivery, so a quiet channel stays quiet unless a rule sends something to it.

Chat. Mention the agent in a channel and it investigates, then replies in the thread. Direct messages to the bot work too. The reply includes a View in Sazabi link to the same thread in the dashboard.

Connecting Slack also gives the agent Slack's MCP tools automatically. You can optionally link your personal Slack account under connected accounts so the agent acts as you.

Questions it answers

• Which issues landed in this channel today, and which are still open?
• Why are checkout errors elevated in the last hour?
• What came back after a resolved issue was re-opened?
• Which delivery rule sent this card to this channel, and what severity does it match?
• What does the agent say about this error when I mention it in a thread?

A worked example: severity-routed cards and a follow-up question

You want only serious issues in #prod-incidents. The docs show this rule shape for the CLI:

sazabi notifications delivery-rules create \
  --notification-type issue_triggered \
  --destination slack_channel:C0123EXAMPLE \
  --severity high \
  --severity critical

Severity conditions match exactly, so selecting high alone would not include critical. That is why the rule lists both.

When an issue opens, the channel sees something like this. The values are made up.

#prod-incidents
Sazabi APP   Issue triggered: checkout-api payment confirmations failing
             Severity: high    [Resolve] [Ignore] [Mute] [Re-open]

priya        @Sazabi why are checkout errors elevated in the last hour?
Sazabi       (thread reply with findings)
             View in Sazabi

Priya's mention starts a thread. The agent searches the project's data and answers under her message. If she needs the full evidence trail, the link opens the same thread in the dashboard. Later, someone clicks Resolve on the card, and the issue closes in Sazabi too.

Slack's own platform shows why the bot has to be in the channel first. Slack's app_mention event requires the app_mentions.read scope and a bot user, and it is not sent for direct messages, which use message.im. Buttons live in section or actions blocks, and an app must acknowledge the resulting block_actions payload within 3 seconds. Sazabi's docs say the bot must be added to the channel, and that missing cards usually mean the rule or the bot membership is wrong.

How it fits with other integrations

Slack is one of several shared destinations, beside Microsoft Teams for chat and PagerDuty for paging. Many teams send high-severity issues to both a channel and an on-call tool. When a fix needs tracking, Linear holds the ticket. See the integrations index for every method, and pricing for plan details.