> ## Documentation Index
> Fetch the complete documentation index at: https://docs.traversal.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Traversal Workers
> Proactive teammates that help you get the most out of Traversal.
Traversal Workers are in **beta**, enabled per team — [read the launch announcement](https://www.traversal.com/blog/ai-sre-proactive-incident-response-traversal-workers). Talk to your Traversal contact to turn them on for your workspace.
**Traversal Workers** are a new and more capable way to work with Traversal. Rather than waiting for you to reach for Traversal, a Worker comes to you — it joins a channel and works alongside your team like a teammate, reading the situation and doing what's worth doing on its own. You never have to know it's there, catch the right moment, or learn how to prompt it — there's no wrong way to use it. It just works.
Under the hood, a Traversal Worker runs on the same Causal Search Engine and Production World Model that power every Traversal investigation. What's had the big upgrade is the experience of working with Traversal across the different use cases that matter most — Workers make that intelligence effortless to consume, meeting you in the flow of the work instead of waiting to be summoned.
## During an incident
When an incident channel is created, a Worker joins and posts a short announcement so the room knows it's there.
Before weighing in, a Worker actively gathers the context it needs to start from the right place. When a channel begins sparse — no description, no messages yet — it searches out the relevant signals rather than guessing, so it may take a minute or two to get its bearings instead of running off in the wrong direction.
From there it works the incident like a good teammate would:
* **Proactive by design.** You don't summon it or schedule it. When someone poses a question or drops a key detail in the channel, a Worker can pick it up and run with it — even if they didn't know Traversal was there.
* **Always listening.** It takes in everything happening across the channel as it arrives — not on a timer, and not by matching keywords, but by continuously reading the conversation the way a teammate would, so it acts at the right moment and never on stale context.
* **The right etiquette.** We've put a lot of work into making a Worker feel like a real teammate — knowing when to open a thread versus post to the room, when a quick emoji or a one-line acknowledgment is enough, when to speak up, and when to stay out of the way. It reads the room instead of adding noise.
* **Fast.** By reusing investigative work it has already done, its responses often land in seconds.
* **Efficient.** Model caching and disciplined use of inference keep it responsive and cost-flat, even across a long, multi-hour or multi-day incident.
## When the incident resolves
Because a Worker stays in sync with the incident the whole way through, it knows when the incident is over. As soon as the room reaches consensus that it's resolved, it drops a **draft post-mortem** right in the channel — reconciling its own investigation with everything the team said and did — and states plainly why it's posting, so no one is left wondering.
See [Post-mortems](/using-traversal/post-mortems) for more details.
## Staying in control
* **Stop proactive mode.** The announcement a Worker posts on arrival includes a one-click button to remove it from the channel — no need to DM anyone or dig through settings.
* **Restraint by design.** A Worker holds back by default and posts only evidence-backed, incident-relevant messages. No rapport-building, no commentary, no guessing from chat alone.
* **Least privilege.** Workers operate within the same Slack scopes described on the [Slack](/using-traversal/slack#permissions) page — read-only where they can be, and they never change your alerting, paging, or routing.
## Getting started
Workers are enabled per team during the beta. Once turned on, a Worker automatically joins qualifying incident channels — there's nothing to configure per incident and nothing to maintain.
The more Traversal knows about your systems, the sharper a Worker's findings. Pair Workers with [Knowledge Bank](/using-traversal/knowledge-bank) so it knows your services, runbooks, and how your team prioritizes.
Today, Workers run in **Slack**. Support for **Microsoft Teams** is coming soon.
## FAQ
Tagging `@Traversal` and [triggers](/using-traversal/slack#triggers) are things you invoke or configure ahead of time. A Worker is proactive: it decides for itself when to engage during a live incident, investigates without being asked, and stays for the whole thing rather than handing back a single report.
No. A Worker errs toward silence and posts only when it has something evidence-backed and incident-relevant to say. Replies thread by default; the main channel is reserved for findings that the whole room needs.
Yes. Click **Stop proactive mode** on the announcement it posts when it joins, and it leaves the channel immediately.
No. A Worker investigates, explains, and drafts a postmortem. It never modifies your alerting, paging, or routing — you stay in control of what happens next.
## Support
Questions or issues? Email [support@traversal.com](mailto:support@traversal.com).