intergral · CandiceChislett · Apr 15, 2026 · Apr 15, 2026 · Apr 15, 2026 · Apr 15, 2026
diff --git a/docs/Admin-and-data/Catalog/catalog.md b/docs/Admin-and-data/Catalog/catalog.md
@@ -2,6 +2,8 @@
 
 The Service Catalog is a persistent record of everything you run - your services and the infrastructure they depend on. Unlike the Services overview, which only shows what is currently sending telemetry, the catalog tracks entries whether they are active or quiet. It gives every service an owner, a tier, and a home, so when something breaks at 2am, nobody has to guess who to contact or what is affected.
 
+The catalog is also the brain behind Coworker. When Coworker investigates an alert or runs a check, it draws on catalog entries to understand what a service does, who owns it, what it depends on, and what has happened to it before. That pre-loaded context means Coworker spends fewer AI Tokens rediscovering information it could already know - and produces better investigations as a result. The more complete your catalog, the more effective Coworker becomes.
+
 The catalog also powers incident response: affected services on an incident link directly to their catalog entries, runbooks surface based on catalog service matches, and Analytics surfaces which catalog entries appear most frequently in incidents.
 
 ![!Screenshot](../../Data-insights/Features/Incidents/images/catalog-list.png)

diff --git a/docs/Coworker/budget-settings.png b/docs/Coworker/budget-settings.png
diff --git a/docs/Coworker/change-what-see.png b/docs/Coworker/change-what-see.png
diff --git a/docs/Coworker/cost-optimize.png b/docs/Coworker/cost-optimize.png
diff --git a/docs/Coworker/coworker-activity.png b/docs/Coworker/coworker-activity.png
diff --git a/docs/Coworker/coworker-behaviour.png b/docs/Coworker/coworker-behaviour.png
diff --git a/docs/Coworker/coworker.png b/docs/Coworker/coworker.png
diff --git a/docs/Coworker/create-event-source.png b/docs/Coworker/create-event-source.png
diff --git a/docs/Coworker/create-event.png b/docs/Coworker/create-event.png
diff --git a/docs/Coworker/create-task.png b/docs/Coworker/create-task.png
diff --git a/docs/Coworker/dashboard.png b/docs/Coworker/dashboard.png
diff --git a/docs/Coworker/getting-started.png b/docs/Coworker/getting-started.png
diff --git a/docs/Coworker/hide-insights.png b/docs/Coworker/hide-insights.png
diff --git a/docs/Coworker/hide-similar.png b/docs/Coworker/hide-similar.png
diff --git a/docs/Coworker/knowledge.gif b/docs/Coworker/knowledge.gif
diff --git a/docs/Coworker/knowledge.png b/docs/Coworker/knowledge.png
diff --git a/docs/Coworker/looking-good.png b/docs/Coworker/looking-good.png
diff --git a/docs/Coworker/new-thread.png b/docs/Coworker/new-thread.png
diff --git a/docs/Coworker/onboarding1.png b/docs/Coworker/onboarding1.png
diff --git a/docs/Coworker/onboarding2.png b/docs/Coworker/onboarding2.png
diff --git a/docs/Coworker/onboarding3.png b/docs/Coworker/onboarding3.png
diff --git a/docs/Coworker/onboarding4.png b/docs/Coworker/onboarding4.png
diff --git a/docs/Coworker/op-alerts.png b/docs/Coworker/op-alerts.png
diff --git a/docs/Coworker/optimize-accept.png b/docs/Coworker/optimize-accept.png
diff --git a/docs/Coworker/optimize-alert.png b/docs/Coworker/optimize-alert.png
diff --git a/docs/Coworker/optimize.png b/docs/Coworker/optimize.png
diff --git a/docs/Coworker/overview1.png b/docs/Coworker/overview1.png
diff --git a/docs/Coworker/per-task-usage.png b/docs/Coworker/per-task-usage.png
diff --git a/docs/Coworker/preferences.png b/docs/Coworker/preferences.png
diff --git a/docs/Coworker/rerun-onboarding.png b/docs/Coworker/rerun-onboarding.png
diff --git a/docs/Coworker/restart-chat.png b/docs/Coworker/restart-chat.png
diff --git a/docs/Coworker/scheduled-task.png b/docs/Coworker/scheduled-task.png
diff --git a/docs/Coworker/settings-budget.png b/docs/Coworker/settings-budget.png
diff --git a/docs/Coworker/triage.png b/docs/Coworker/triage.png
diff --git a/docs/Coworker/usage-ai-token.png b/docs/Coworker/usage-ai-token.png
diff --git a/docs/Coworker/usage-report.png b/docs/Coworker/usage-report.png
diff --git a/docs/Coworker/usage1.png b/docs/Coworker/usage1.png
diff --git a/docs/Data-insights/Features/OpsPilot/Coworker/faq.md b/docs/Data-insights/Features/OpsPilot/Coworker/faq.md
@@ -0,0 +1,183 @@
+# Coworker FAQ
+
+## Getting started
+
+### How do I get started with Coworker?
+
+When you first open Coworker, a [guided setup conversation](overview.md#getting-started) walks you through setting your preferences and creating your first task. The quickest way to get value is to enable [OpsPilot Alerts](overview.md#opspilot-alerts), which connects Coworker to your existing alert rules so it automatically investigates whenever one fires.
+
+### Can I have more than one Coworker?
+
+No. Each user has one Coworker. Your feed and preferences are personal to you, but tasks and investigations are shared across your organisation — everyone on the team can see what Coworker has raised.
+
+### Can my team share a Coworker?
+
+Tasks and investigations are already shared across your organisation. Use the **Just for me** dropdown at the top of the dashboard to toggle between your personalised feed and the full team view.
+
+### Can I restart the setup?
+
+Yes. Click the settings icon on the Coworker dashboard, select **Reset onboarding**, and click **Open onboarding** to walk through the setup flow again. Re-running does not delete anything you already have.
+
+![!Screenshot](../../../../Coworker/rerun-onboarding.png)
+
+---
+
+## Tasks
+
+### How many tasks should I create?
+
+Start with one or two [scheduled tasks](overview.md#scheduled-tasks) covering your most critical services, and enable [OpsPilot Alerts](overview.md#opspilot-alerts). Add more tasks over time as you identify gaps. Too many tasks running frequently can increase AI Token usage.
+
+### How often should I run scheduled tasks?
+
+Daily is the most common cadence and a good starting point. Every 6 hours works well for services that need closer attention. Weekly is often enough for higher-level system reviews.
+
+If you find yourself wanting very frequent checks on a specific pattern, a [monitoring task](tasks.md#monitoring-tasks) is usually a better fit than a high-frequency scheduled task. If something needs near-real-time response, connecting an alert rule via OpsPilot Alerts will be more effective and much cheaper.
+
+!!! info "Learn more"
+    [Scheduled tasks](overview.md#scheduled-tasks)
+
+### Why isn't my task finding anything?
+
+It may take a few runs for Coworker to build enough context to surface meaningful insights. If a task consistently finds nothing, consider adjusting the description to be more specific about what you want it to investigate.
+
+### What is the difference between a scheduled task and a monitoring task?
+
+A [scheduled task](overview.md#scheduled-tasks) runs on a recurring interval and produces a general report of findings. A [monitoring task](overview.md#monitoring-tasks) is focused on a specific pattern or issue (created from an insight) and tracks whether that pattern is improving, worsening, or stable over time.
+
+---
+
+## Insights
+
+### What is the difference between Resolve and Ignore?
+
+**Resolve** marks an insight as handled and it will appear in your resolved insights history. **Ignore** dismisses it from your priority list without marking it as resolved. Use Resolve when you've taken action; use Ignore when the insight isn't relevant to you.
+
+!!! info "Learn more"
+    [Insights](overview.md#insights)
+
+### Why am I seeing the same insight repeatedly?
+
+If the underlying issue hasn't been fixed, Coworker will continue to surface it. The occurrence history on each insight shows whether it is a recurring pattern. Use **Watch** to create a [monitoring task](overview.md#monitoring-tasks) that tracks whether the issue improves.
+
+### How do I change what types of insights I see?
+
+Click **Change what I show you** on the dashboard to adjust your severity and category preferences (Errors, Performance, Notable, Coverage), or use **Update via chat** to describe your preferences in plain language.
+
+!!! info "Learn more"
+    [Preferences](overview.md#preferences)
+
+---
+
+## Costs
+
+### What are OpsPilot AI Tokens?
+
+OpsPilot AI Tokens are the usage allowance for Coworker's AI-powered work, including chat, alert investigations, scheduled checks, telemetry analysis and recommendations. Your plan includes a fixed monthly allowance, and OpsPilot gives you clear usage visibility, forecasting and controls so there are no surprises.
+
+!!! info "Learn more"
+    [Understanding OpsPilot AI Tokens](usage.md#understanding-opspilot-ai-tokens)
+
+### What uses AI Tokens?
+
+AI Tokens are used whenever Coworker performs AI-powered work:
+
+- Answering questions in chat
+- Investigating alerts and situations
+- Triaging and performing background checkups on open situations
+- Analysing telemetry and service behaviour
+- Running scheduled checks
+- Generating recommendations, suggested fixes and debriefs
+- Updating situations and producing findings
+
+### Does every Coworker action use the same number of AI Tokens?
+
+No. Usage depends on the amount of telemetry, context and reasoning required. A simple chat question typically uses fewer AI Tokens than a deeper investigation that reviews metrics, logs, prior findings and service context before generating a recommendation.
+
+### Can I forecast my AI Token usage?
+
+Yes. The **Projected monthly** metric in the Usage view estimates your end-of-month AI Token consumption based on current usage patterns, so you can see whether you are on track to stay within your plan allowance.
+
+!!! info "Learn more"
+    [Activity tab](usage.md#activity-tab)
+
+### How much does Coworker cost to run?
+
+Cost depends on several factors:
+
+- How many tasks you have and how frequently they run
+- The [model tier](tasks.md#model-tier) selected (Thorough uses more AI Tokens than Efficient)
+- The number of events received from your event sources - the more alerts that fire, the more investigations Coworker runs
+- The number of open situations - more open situations means more background checkups running continuously
+
+You can set a monthly task allowance to control spend, with configurable warning and halt thresholds to prevent overruns.
+
+!!! info "Learn more"
+    [Cost and Optimisation](usage.md)
+
+### How do I reduce AI Token usage?
+
+- Review **Optimization Suggestions** in the [AI Tokens tab](usage.md#optimization-suggestions); these appear automatically after a task has run a few times and Coworker detects ways it could be improved
+- **Apply** an optimisation suggestion to apply the recommended change immediately
+- Click **Analyse & Optimise** to trigger an on-demand optimisation review at any time
+- Switch high-volume or routine tasks to the [Efficient model tier](tasks.md#model-tier)
+- Reduce the frequency of scheduled tasks that run often but find little
+- Review the **AI Token Breakdown** table to identify the most expensive tasks and consolidate or adjust them
+- If noisy alerts are driving up costs, consider disabling Coworker from investigating them. Click on the **OpsPilot Alerts** event source in the sidebar and sort by **Most events** to see which alert rules are firing most frequently - those are the best candidates to review or exclude
+
+### What happens when I reach my task allowance?
+
+Coworker will stop running tasks once spend reaches the **Halt threshold**, which by default is set to 100% of your task allowance. You can lower this threshold to stop tasks earlier and protect your allowance. A separate **Warning threshold** notifies you before the halt is reached.
+
+!!! info "Learn more"
+    [Allowance](usage.md#ai-token-allowance)
+
+### Can I get warned before hitting my allowance limit?
+
+Yes. The **Warning threshold** in Settings > Budget & cost triggers a notification when your spend reaches a set percentage of your task allowance (e.g. 80%). This gives you time to adjust tasks or increase your allowance before tasks are halted.
+
+!!! info "Learn more"
+    [Allowance](usage.md#ai-token-allowance)
+
+### How do I accept or dismiss an optimisation suggestion?
+
+Open the **AI Tokens** tab in Usage and scroll to **Optimization Suggestions**. Expand any suggestion to see the reasoning under **Why this suggestion** and the proposed change under **Instruction changes**. Click **Apply** to apply it immediately, or **Dismiss** to ignore it.
+
+!!! info "Learn more"
+    [Optimization Suggestions](usage.md#optimization-suggestions)
+
+### What is the difference between Thorough and Efficient model tiers?
+
+**Thorough** handles any task and is more capable. Use it for critical alerts and complex investigations where depth matters. **Efficient** is suited to simpler, focused tasks and costs less. Use it for routine or high-volume tasks to keep spend down. You can set the model tier per task or per event source.
+
+!!! info "Learn more"
+    [Model tier](tasks.md#model-tier)
+
+### Why use AI Tokens instead of unlimited AI?
+
+AI-powered investigations consume compute and reasoning resources. OpsPilot AI Tokens give your team a predictable, fixed allowance for Coworker's work, with full visibility into what was used and what it delivered. This keeps AI usage transparent and controllable for both teams and budgets.
+
+---
+
+## Memory
+
+### How long does it take for Coworker to become useful?
+
+Coworker starts providing value immediately, but becomes noticeably smarter after a few days of running tasks. As it builds [memory](overview.md#memory) about your services and patterns, its insights become more relevant and its task analysis more accurate.
+
+### Can I clear Coworker's memory?
+
+Please contact support if you need to reset Coworker's [memory](overview.md#memory).
+
+---
+
+## Privacy and data
+
+### What data does Coworker have access to?
+
+Coworker has access to the observability data in your OpsPilot account: metrics, logs, traces, and alert rules. It does not have access to data outside your organisation's account.
+
+---
+
+!!! question "Need more help?"
+    Contact support in the chat bubble and let us know how we can assist.
diff --git a/docs/Data-insights/Features/OpsPilot/Coworker/getting-started.md b/docs/Data-insights/Features/OpsPilot/Coworker/getting-started.md
@@ -0,0 +1,100 @@
+# Getting started with Coworker
+
+## Onboarding
+
+When you first access Coworker, a short guided setup walks you through personalising your experience. You can skip it at any time and still get a working Coworker, though you'll see a broader, less tailored view until you configure your preferences.
+
+### Step 1: Introduction
+
+Step 1 introduces what Coworker does and how it works, with four cards showing the flow: something fires, Coworker investigates, you get a clear story, and it keeps checking back. Click **Quick start** to begin, or **Skip** to go straight to the dashboard.
+
+![!Screenshot](../../../../Coworker/onboarding1.png)
+
+---
+
+### Step 2: Your role
+
+![!Screenshot](../../../../Coworker/onboarding2.png)
+
+Coworker asks which role best describes you. This sets the defaults for which kinds of situations are surfaced to you first:
+
+| Role | Description |
+|---|---|
+| **Developer** | Building features, owning services |
+| **SRE / DevOps** | On-call, infra, reliability |
+| **Tech lead** | Leading a team, mostly hands-on |
+| **Manager / leadership** | Team or org-level visibility |
+| **Something else…** | Tell me in your own words |
+
+Pick the closest fit. Select **Skip - I'll come back later** to proceed without setting a role.
+
+---
+
+### Step 3: What do you want me to help with?
+
+![!Screenshot](../../../../Coworker/onboarding3.png)
+
+Select as many as apply:
+
+| Option | Description |
+|---|---|
+| **Surface critical issues** | Stay quiet most of the time; get loud when something's actually breaking |
+| **Filter alert noise** | I get a lot of alerts already - only show me the ones that matter |
+| **Catch cascading issues** | Spot when several small problems are really one big one across services |
+| **Run regular checks** | Watch specific services or metrics on a cadence and flag what looks off |
+| **Give me regular reports for the team** | Uptime, SLO burn, incident counts, deploy frequency - the kinds of numbers you have to share up |
+| **Verify deploys** | After each deploy, check things look healthy and tell me if anything regressed |
+| **Brief me on my shift** | Tell me what happened while I was away, what's still open, what's resolved |
+| **Take action on routine stuff** | Run scripts, kick off jobs, hit endpoints - automate the small stuff I'd otherwise context-switch for |
+
+Click **Continue** when done, or **Skip** to proceed without selecting any.
+
+---
+
+### Step 4: Which of these matter most to you?
+
+![!Screenshot](../../../../Coworker/onboarding4.png)
+
+The final step. Select the domains you want Coworker to prioritise in your feed:
+
+| Area | What it covers |
+|---|---|
+| **Errors and exceptions** | Crashes, failed requests, exception spikes |
+| **Application performance** | Slow endpoints, latency spikes, throughput dips |
+| **Infrastructure and runtime** | Hosts, containers, pods, CPU / memory / disk |
+| **Databases and data stores** | Queries, connections, replication, cache health |
+| **Data pipelines and quality** | DAG failures, freshness, schema drift, model output |
+| **Deploys and releases** | Rollouts, regressions, config changes |
+| **Team and delivery health** | Cross-service stability, release confidence, incident rate |
+| **Reliability and SLOs** | Burn rate, availability, alert noise |
+| **Cost and capacity** | Spend anomalies, quota limits, overprovisioning |
+| **Security and auth** | Auth failures, suspicious traffic, IAM, secrets |
+
+Select as many as apply. These can be updated at any time from **Settings > Your preferences**.
+
+---
+
+## Configuring your view
+
+Whether set during onboarding or later from Settings, three controls decide what appears in your Coworker feed:
+
+| Setting | Description |
+|---|---|
+| **Focus services** | The specific services you own or care about, by name or glob pattern (e.g. `payments-*`). Situations touching these are prioritised in your feed. |
+| **Focus areas** | The domains you want prioritised: Errors and exceptions, Application performance, Infrastructure and runtime, Databases and data stores, Data pipelines and quality, Deploys and releases, Team and delivery health, Reliability and SLOs, Cost and capacity, Security and auth. |
+| **Custom keywords** | Any terms beyond the focus areas above - a library, technology, or feature name specific to your stack. A match nudges related situations into your feed. |
+
+None of these change what Coworker investigates or raises across your organisation. They only change what reaches your personal feed.
+
+---
+
+## Settings
+
+Open **Settings** from the Coworker dashboard to access your preferences, check-in cadence, behaviour, and budget controls. You don't need to configure all of this up front — sensible defaults are in place, and you can ask Coworker directly in any chat to update your preferences conversationally.
+
+See [Settings](settings.md) for full details.
+
+---
+
+!!! question "Need more help?"
+    Contact support in the chat bubble and let us know how we can assist.
diff --git a/docs/Data-insights/Features/OpsPilot/Coworker/knowledge.md b/docs/Data-insights/Features/OpsPilot/Coworker/knowledge.md
@@ -0,0 +1,51 @@
+# Knowledge
+
+Coworker builds a growing understanding of your systems, your team, and your preferences over time. Everything it does - investigating alerts, running tasks, talking to you - adds to this knowledge and carries forward into future investigations.
+
+Click the **Knowledge** button on the Coworker dashboard to explore what Coworker has learned.
+
+The header shows a live count of total memories, observations, nodes, and links across your organisation.
+
+## Memory types
+
+| Memory type | What it holds |
+|---|---|
+| **System-wide** | How your services fit together, what's normal, and what tends to break. Shared across your whole organisation, so what Coworker learns helping one person makes it smarter for everyone |
+| **Task-specific** | What recurring checks have turned up before and the patterns that matter. Can reduce token costs by up to 50% on long-running tasks |
+| **Team** | Who owns what, where the runbook lives, what each channel is for |
+| **User** | Your personal preferences and the way you like to work, learned from your conversations |
+
+---
+
+## Graph view
+
+A visual map of everything Coworker knows. Nodes represent entities (services, databases, concepts) and links between them show relationships. Node size reflects how frequently an entity is referenced.
+
+![!Screenshot](../../../../Coworker/knowledge.gif)
+
+The **Entities** panel on the right lists every entity ranked by reference count. Use the search box to find a specific one.
+
+---
+
+## List view
+
+A searchable, filterable list of individual memories.
+
+| Control | Description |
+|---|---|
+| **Search memories** | Find specific facts by keyword |
+| **Filter by entity** | Scope the list to a particular service or concept |
+| **Sort** | Order by newest or oldest |
+
+Each memory shows the fact Coworker recorded, when it was added, and any entity tags attached to it.
+
+---
+
+## Correcting Coworker
+
+When Coworker raises something that isn't a problem, dismiss it with a quick reason, such as "this is expected" or "too noisy". Coworker turns your correction into a lasting fact: next time it sees the same pattern on the same service, it remembers and won't raise it again. A few early corrections go a long way towards tuning Coworker to your reality.
+
+---
+
+!!! question "Need more help?"
+    Contact support in the chat bubble and let us know how we can assist.