Using Traversal in Slack

With Traversal's AI SRE, engineers can investigate incidents, triage alerts, and explore system behavior directly from Slack.

This page explains how to trigger Traversal, how investigations appear, and how to configure automatic workflows for your alert channels.


Installing Slack

Before you get started, ensure that you have the Slack app installed.

  1. Log in to Traversal at app.traversal.comarrow-up-right

  2. Go to Settings → Integrations to access the Integration Manager

  3. Find the Slack integration within Productivity integrations

  1. Input your Slack bot token for API access

  2. You'll be redirected to Slack to install/authorize the Traversal app in your workspace.

  3. Choose the workspace and approve the requested permissions.

  4. Back in Traversal, click Test connection.

  5. Click Save.

What You Can Do with Traversal in Slack

With the Traversal Slack integration, engineers can:

  • Start an investigation from any channel or DM — either manually or via custom auto-triggers

  • See Traversal investigations for incidents or alerts shared directly in their Slack channels

  • Ask Traversal follow-up questions in natural language to refine or extend the analysis

  • Run RCA or Alert Triage instantly by @Traversal

  • Control when Traversal responds automatically using /register-triggerrules


Triggering Traversal Manually

You can invoke Traversal at any time by mentioning it in a message. For example:

@Traversal why are we seeing no requests on checkout?

Once you @Traversal, Traversal will automatically:

  • Collect relevant context from the Slack thread, if applicable

  • Run an RCA (deep mode) or Chat (fast mode) investigation, depending on your message

  • Analyze logs, metrics, traces, and alerts

Return hypotheses, evidence, anomalies, and impact analysis.


Using /register-trigger to Configure Automatic Investigations

Traversal can automatically investigate alerts as they appear in Slack channels — without manual tagging.

Use the command: /register-trigger

Examples of ways to configure rules:

By Sender

  • Datadog bot

  • Grafana bot

  • PagerDuty bot (Prevents Traversal from triggering on human messages)

By Severity

  • Critical / Sev1 / ErrorRateHigh

  • Custom severity naming conventions

Product / Service / Tag Patterns

  • [checkout]

  • [billing]

  • [core]

Metric or Log Patterns

  • CPU saturation

  • WAF blocks

  • webhook failures

  • latency spikes

Data Source

  • Datadog

  • Grafana

  • PagerDuty

Common Trigger Configurations

Here are patterns most pilot customers adopt:

  • Bot-only trigger: Only fire when an alert originates from Datadog/Grafana/PagerDuty.

  • Severity filter: Only investigate sev1 or critical alerts.

  • Service or system filter: Only trigger Traversal for alerts tagged [important-product].

  • Alert-type filter: Investigate alerts containing terms like “latency”, “error”, “callback”, “anomaly”.


How to Ask Follow-Up Questions

You can deepen or refine an investigation by asking follow-ups. Examples below:

@Traversal show me impacted services

@Traversal compare this incident to last week’s checkout outage

@Traversal re-run RCA with today’s context

Traversal maintains context within the Slack thread, allowing iterative investigation.

circle-info

You can move and view the Slack session to the Traversal web app at any point and continue asking follow-ups there.

Last updated