> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/NVIDIA/NemoClaw/llms.txt
> Use this file to discover all available pages before exploring further.

# Customize the network policy

> Add, remove, or modify allowed endpoints in the sandbox policy. Covers static changes, dynamic updates, and policy presets.

# Customize the network policy

The NemoClaw sandbox operates under a strict-by-default network policy. Only endpoints listed in `nemoclaw-blueprint/policies/openclaw-sandbox.yaml` are allowed by default. You can extend this policy through static changes to the policy file, dynamic updates to a running sandbox, or by applying a preset for a common service.

## Prerequisites

* The NemoClaw source repository for static changes.
* A running NemoClaw sandbox for dynamic changes.
* The OpenShell CLI on your `PATH`.

## Static changes

Static changes modify the baseline policy file. They take effect the next time the sandbox is created or re-onboarded.

<Steps>
  <Step title="Edit the policy file">
    Open `nemoclaw-blueprint/policies/openclaw-sandbox.yaml`. Each entry in the `network_policies` section defines an endpoint group.

    The policy file uses the following structure:

    ```yaml theme={null}
    network_policies:
      my_service:
        name: my_service
        endpoints:
          - host: api.example.com
            port: 443
            protocol: rest
            enforcement: enforce
            tls: terminate
            rules:
              - allow: { method: GET, path: "/**" }
              - allow: { method: POST, path: "/api/**" }
        binaries:
          - { path: /usr/local/bin/openclaw }
    ```

    The fields are:

    * **`endpoints`** — Host, port, and HTTP rules for each allowed destination.
    * **`binaries`** — Executables permitted to use this endpoint group.
    * **`rules`** — HTTP methods and path patterns that are permitted.
  </Step>

  <Step title="Re-run onboard">
    Apply the updated policy by re-running the onboard wizard:

    ```bash theme={null}
    nemoclaw onboard
    ```

    The wizard picks up the modified policy file and applies it to the sandbox.
  </Step>

  <Step title="Verify the policy">
    Confirm the sandbox is running with the new configuration:

    ```bash theme={null}
    openclaw nemoclaw status
    ```
  </Step>
</Steps>

## Dynamic changes

Dynamic changes apply a policy update to a running sandbox without restarting it. They are useful for testing or for temporarily allowing access to a new endpoint.

<Steps>
  <Step title="Create a policy file">
    Write a YAML file with the additional endpoints. Use the same format as the baseline policy:

    ```yaml theme={null}
    network_policies:
      pypi:
        name: pypi
        endpoints:
          - host: pypi.org
            port: 443
            protocol: rest
            enforcement: enforce
            tls: terminate
            rules:
              - allow: { method: GET, path: "/**" }
          - host: files.pythonhosted.org
            port: 443
            rules:
              - allow: { method: GET, path: "/**" }
        binaries:
          - { path: /usr/bin/python3 }
          - { path: /usr/local/bin/pip }
    ```
  </Step>

  <Step title="Apply the policy">
    Push the policy update to the running sandbox:

    ```bash theme={null}
    openshell policy set ./my-policy.yaml
    ```

    The change takes effect immediately — no sandbox restart is required.
  </Step>
</Steps>

<Warning>
  Dynamic changes apply only to the current session. When the sandbox stops, the running policy resets to the baseline defined in `openclaw-sandbox.yaml`. To make changes permanent, edit the static policy file and re-run `nemoclaw onboard`.
</Warning>

## Policy presets

The `nemoclaw-blueprint/policies/presets/` directory contains ready-to-use policy files for common services. Apply a preset dynamically with `openshell policy set`:

<Tabs>
  <Tab title="Slack">
    Allows the Slack API, webhooks, and API subdomain on port 443:

    ```bash theme={null}
    openshell policy set nemoclaw-blueprint/policies/presets/slack.yaml
    ```

    Endpoints: `slack.com`, `api.slack.com`, `hooks.slack.com`
  </Tab>

  <Tab title="Hugging Face">
    Allows the Hugging Face Hub, LFS CDN, and inference API:

    ```bash theme={null}
    openshell policy set nemoclaw-blueprint/policies/presets/huggingface.yaml
    ```

    Endpoints: `huggingface.co`, `cdn-lfs.huggingface.co`, `api-inference.huggingface.co`
  </Tab>

  <Tab title="Other presets">
    The following presets are available in `nemoclaw-blueprint/policies/presets/`:

    | Preset file        | Service                        |
    | ------------------ | ------------------------------ |
    | `discord.yaml`     | Discord API                    |
    | `docker.yaml`      | Docker Hub registry            |
    | `huggingface.yaml` | Hugging Face Hub and inference |
    | `jira.yaml`        | Jira REST API                  |
    | `npm.yaml`         | npm registry                   |
    | `outlook.yaml`     | Microsoft Outlook / Graph API  |
    | `pypi.yaml`        | Python Package Index           |
    | `slack.yaml`       | Slack API and webhooks         |
    | `telegram.yaml`    | Telegram Bot API               |

    Apply any preset:

    ```bash theme={null}
    openshell policy set nemoclaw-blueprint/policies/presets/<preset>.yaml
    ```
  </Tab>
</Tabs>

To make a preset permanent, copy the `network_policies` block from the preset file into `openclaw-sandbox.yaml` and re-run `nemoclaw onboard`.

## Policy file field reference

| Field                              | Type    | Description                                                  |
| ---------------------------------- | ------- | ------------------------------------------------------------ |
| `name`                             | string  | Unique identifier for this policy group                      |
| `endpoints[].host`                 | string  | Hostname to allow                                            |
| `endpoints[].port`                 | integer | Port number (typically `443`)                                |
| `endpoints[].protocol`             | string  | Protocol type, e.g. `rest`                                   |
| `endpoints[].enforcement`          | string  | `enforce` to block violations, `audit` to log only           |
| `endpoints[].tls`                  | string  | `terminate` to inspect TLS traffic                           |
| `endpoints[].rules[].allow.method` | string  | HTTP method: `GET`, `POST`, `PUT`, `PATCH`, `DELETE`, or `*` |
| `endpoints[].rules[].allow.path`   | string  | Path glob, e.g. `/**` or `/api/**`                           |
| `binaries[].path`                  | string  | Absolute path to the binary allowed to use these endpoints   |

## Related topics

<CardGroup cols={2}>
  <Card title="Approve network requests" icon="check-circle" href="/network-policy/approve-network-requests">
    Use the TUI to allow or block requests from the running agent in real time.
  </Card>

  <Card title="Network policies reference" icon="book" href="/reference/network-policies">
    Baseline policy listing, filesystem rules, and the full operator approval flow.
  </Card>
</CardGroup>