---
title: "Cloud Agents"
description: "On-demand and scheduled AI agents for Kubernetes analysis, cost optimization, security auditing, and remediation — launched directly from Nirmata Control Hub."
diataxis: how-to
applies_to:
  product: "nirmata-ai-agents"
audience: ["platform-engineer"]
last_updated: 2026-04-16
url: https://docs.nirmata.io/docs/control-hub/agent-hub/cloud-agents/
---


> **Applies to:** Nirmata AI Agents 1.0 and later

Cloud Agents are AI-powered workflows you launch on-demand or on a schedule from Nirmata Control Hub. Each agent runs as a short-lived job, executes a specific analysis or remediation task, and produces a detailed report — with no persistent infrastructure required.

For a conceptual overview of what Cloud Agents are and how they work, see [Cloud Agents in Nirmata AI Agents](/docs/ai/cloud-agents/).

## In This Section

- [**Agent Catalog**]({{< relref "agent-catalog/" >}}) — Browse available agents, see what each one does, and find the right agent for your task
- [**Launching an Agent**]({{< relref "agent-launch/" >}}) — Run any agent on-demand against a target cluster
- [**Schedules**]({{< relref "schedules/" >}}) — Configure recurring agent runs with notifications
- [**Agent Runs**]({{< relref "agent-runs/" >}}) — Monitor active runs, view live logs, and access completed reports

## Accessing Cloud Agents

Navigate to **Agent Hub → Cloud Agents** in Nirmata Control Hub.


---

## Agent Catalog


The Agent Catalog is the directory of all Cloud Agents available in your Nirmata Control Hub instance. Navigate to **Agent Hub → Cloud Agents → Agent Catalog** to access it.

## What the Catalog Shows

Each agent entry in the catalog displays:

- **Name and description** — what the agent does and what problem it solves
- **Agent type** — Workflow (multi-step) or Task (single operation)
- **Required inputs** — parameters you must provide at launch time (e.g., cluster, namespace, resource name)
- **Required permissions** — the Kubernetes RBAC permissions the agent needs on the target cluster
- **Recent runs** — a summary of the most recent executions of that agent, including status and when they ran

## Browsing and Searching

You can filter the catalog by:

- **Category** — cost, security, compliance, policy, troubleshooting, remediation
- **Type** — Workflow or Task
- **Target** — agents that run against a cluster, a namespace, or a cloud account

## From the Catalog

From any agent entry you can:

- [**Launch the agent**](../agent-launch/) — run it immediately with the inputs you provide
- [**Schedule the agent**](../schedules/) — set up a recurring run on a cron schedule
- View the **run history** for that agent across all clusters


---

## Launching an Agent


You can launch any Cloud Agent on-demand from the [Agent Catalog](../agent-catalog/). The agent starts immediately, executes against the target you specify, and produces a report when it finishes.

## Launch Steps

1. Open **Agent Hub → Cloud Agents → Agent Catalog**.
2. Find the agent you want to run and click **Launch**.
3. Fill in the launch form:
   - **Target cluster** — required for agents that run against a cluster. The dropdown shows all clusters connected to your Control Hub instance.
   - **Namespace** — required for agents scoped to a namespace (shown only when applicable).
   - **Additional inputs** — agent-specific parameters such as resource names, severity thresholds, or output format. Each field includes a description of what it expects.
4. Click **Run** to start the agent.

## After Launch

Once launched, the run appears immediately in [Agent Runs](../agent-runs/) with status **Running**. You can:

- Follow **live execution logs** and step-by-step progress (for Workflow agents).
- Wait for the run to complete and then view or download the **report**.

## Disconnected Clusters

If the selected cluster is disconnected at launch time, the run will retry automatically before being marked as **Failed**. Check cluster connectivity under **Cluster Management** if a run is stuck retrying.

## Permissions

Some agents require specific Kubernetes RBAC permissions on the target cluster. The Agent Catalog entry for each agent lists its required permissions. Verify these are in place before launching — the agent runs under a scoped service account that is provisioned automatically and cleaned up after the run completes.


---

## Schedules


Schedules let you run any Cloud Agent automatically on a recurring basis. Navigate to **Agent Hub → Cloud Agents → Schedules** to manage them.

## Creating a Schedule

1. Click **New Schedule** (or click **Schedule** on an agent from the [Agent Catalog](../agent-catalog/)).
2. Configure the schedule:
   - **Agent** — select the agent to run.
   - **Target cluster** — the cluster to run the agent against.
   - **Inputs** — the same agent-specific parameters as a [manual launch](../agent-launch/).
   - **Interval** — choose from: Hourly, Every 12 hours, Daily, Weekly, or enter a **custom cron expression**.
   - **Timezone** — the timezone used to interpret the schedule interval.
   - **Enabled** — toggle on to activate immediately, or leave off to save it inactive.
3. Optionally configure **Notifications** (see below).
4. Click **Save**.

## Managing Schedules

From the Schedules list you can:

- **Filter** schedules by agent name, target cluster, or status (active / inactive).
- **Enable or disable** a schedule without deleting it.
- **Edit** an existing schedule to change its interval, inputs, or notification settings.
- **Delete** a schedule you no longer need.
- **View run history** — see all past runs triggered by a specific schedule, with their status and reports.

## Notifications

For each schedule you can configure notifications sent when a scheduled run completes:

| Channel | Setup |
|---------|-------|
| **Email** | Enter one or more recipient addresses. A summary is sent on run completion. |
| **Slack** | Select a Slack channel. Requires a Slack workspace integration configured under **Settings → Integrations**. |

Notifications are sent regardless of whether the run succeeded or failed.

## Cron Expression Reference

For custom intervals, use standard cron syntax (5 fields: minute, hour, day-of-month, month, day-of-week):

```bash


---

## Agent Runs


Agent Runs is the unified view of all Cloud Agent executions — across every agent, every cluster, and every trigger source. Navigate to **Agent Hub → Cloud Agents → Agent Runs**.

## What You Can Do

### Filter and Find Runs

Filter the runs list by:

- **Status** — Running, Completed, Failed, Cancelled
- **Agent name** — view all runs for a specific agent
- **Trigger source** — Manual, Schedule, Copilot, Webhook, or Event
- **Cluster** — runs targeting a specific cluster
- **Date range** — narrow to a specific time window

### View Run Details

Click any run to open its detail view:

| Field | Description |
|-------|-------------|
| **Status** | Current state — Running, Completed, Failed, or Cancelled |
| **Triggered by** | Who or what started the run — a user name, schedule name, Copilot session, or webhook |
| **Target cluster** | The cluster the agent ran against |
| **Started / Completed** | Timestamps for when the run began and ended |
| **Duration** | Total execution time |
| **Inputs** | The parameters that were passed to the agent at launch |
| **Error details** | If the run failed, the error message and step where it failed |

### Follow Live Progress

For runs that are currently in progress, the detail view shows:

- **Live execution logs** — streamed in real time as the agent runs
- **Step-by-step progress** (for Workflow agents) — each step's status, duration, and output as it completes

### View and Share Reports

When a run completes successfully it produces a structured report. From the run detail view you can:

- **View the report** inline
- **Share** a link to the report with teammates
- **Download** the report as a file

### Delete Runs

Select one or more runs and click **Delete** to remove them from the history. Deleting a run also removes its associated report.

## Run Statuses

| Status | Meaning |
|--------|---------|
| **Running** | The agent is currently executing |
| **Completed** | The run finished successfully and a report is available |
| **Failed** | The run encountered an error; check the error details on the run |
| **Cancelled** | The run was manually cancelled before completing |
| **Retrying** | The run failed to reach the target cluster and is retrying automatically |