Add docs for subagents

docs.json

@@ -85,7 +85,8 @@
           "paid/sso-authentication",
           "paid/organization-template-catalog",
           "paid/customer-workspaces",
-          "paid/slack-app"
+          "paid/slack-bot",
+          "paid/subagents"
         ]
       },
       {

paid/sso-authentication.mdx

@@ -1,7 +1,7 @@
 ---
 title: 'SSO Authentication'
 description: "How to enable SSO Authentication in your environment"
-icon: 'grid'
+icon: 'user'
 ---
 
 import { NarrowImage } from '/snippets/narrow-image.jsx'

paid/subagents.mdx

@@ -0,0 +1,44 @@
+---
+title: 'Subagents'
+description: 'AI-powered agents that investigate and analyze your workflows'
+icon: 'sparkles'
+---
+
+<Info>
+**Subagents** is an Enterprise edition feature. See [pricing](https://www.openops.com/pricing/) for details.
+</Info>
+
+Subagents are specialized AI agents that run in isolated environments to handle complex, long-running analysis tasks on your behalf. Unlike the general-purpose AI assistant, each subagent is purpose-built for a specific domain and can work through large amounts of data autonomously before returning a structured result.
+
+![Subagent Page](/images/paid/subagent.png)
+
+## Investigation subagent
+
+The first available subagent is the **Investigation** agent. It analyzes workflow execution data to help you understand why a workflow run failed, behaved unexpectedly, or performed poorly.
+
+The investigation subagent can help you:
+
+- **Diagnose failures** — identify which step failed, why it failed, and what the error means.
+- **Trace unexpected behavior** — follow data through each step to explain why a workflow produced an unexpected result.
+- **Spot performance issues** — find bottlenecks or steps that took longer than expected.
+
+After completing its analysis, the agent streams a structured report with findings, and a root cause assessment.
+
+## Using the investigation subagent
+
+The investigation subagent is available as a tool inside the OpenOps AI chat interface.
+
+To start an investigation:
+
+1. Open the AI chat in your OpenOps environment.
+2. Describe the issue you want to investigate — for example: *"Why did this workflow fail?"* or *"The last run produced unexpected output in the final step, can you check what happened?"*
+3. Provide the **workflow run ID** you want to analyze.
+4. The agent runs in the background and streams progress updates as it works. When complete, it returns its findings directly in the chat.
+
+## How it works
+
+Subagents run in isolation from the rest of your environment, with no persistent storage and no access to external systems beyond the data they are given.
+
+## Enabling subagents for your environment
+
+To learn more or to enable subagents for your environment, contact your OpenOps account manager.