---
sourceDocument: Brazil Build or modify applications
sourceDocumentLink: https://www.servicenow.com/docs/r/application-development

 Release :

    - brazil

ft:locale :

    - en-US

ft:publication_title :

    - Brazil Build or modify applications

ft:clusterId :

    - cadev

bundleId :

    - cadev

workflow :

    - Development, Data, and Analytics


---

# Action approval flow in ServiceNow Arc

# Action approval flow in ServiceNow Arc {#ariaid-title1}

Release version: Brazil  
Updated September 25, 2026  
![](https://www.servicenow.com/docs/portal-asset/ico-clock) 5 minutes to read
Summarize  
![AI sparkle icon](https://servicenow.com/docs/portal-asset/ai-sparkle-icon) Summarized using AI  
This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.  

## Summary of Action approval flow in ServiceNow Arc

The action approval flow in ServiceNow Arc controls whether an action runs immediately, requires user approval, or is blocked.
This flow uses three main controls:tool gates,approval patterns, andauto approval.
These controls help ensure that potentially risky or impactful commands are properly reviewed before execution, protecting your environment and maintaining operational governance.
Show full answer Show less  

## How the Flow Works

Every action in Arc is evaluated in this order:

* Arc checks for an active policy; if none exists, the action is refused.
* Arc applies the tool gate setting for the tool involved, which determines the initial handling:
  * **Allow:** action runs immediately without approval.
  * **Deny:** action is blocked outright.
  * **Human in the loop (HITL):** action is checked against approval patterns.
  * **Auto:** action is reviewed by an AI model with some prompts for hard gates.
* If under HITL, Arc matches the command against approval patterns to decide if user approval is needed.
* If under Auto, the AI model reviewer approves most actions unless the action is under a hard gate or escalated.
* When prompted, users can approve or deny actions; approved actions run with sandbox, network, connector, and file type restrictions.

## Key Features

### Gate Strengths

Approval patterns assign a gate strength to actions:

* **Hard gate:** Requires user approval every time; used for destructive or irreversible actions.
* **Soft gate:** Requires user approval once, which can be saved for repeated similar actions; used for routine, low-impact actions.

### Tool Gates

Each tool in Arc can be set to one of these modes:

* **Allow:** runs all actions without prompts; suited for low-impact, read-only tools.
* **Deny:** blocks all actions; used for disallowed tools.
* **Human in the loop (HITL):** routes commands through approval patterns for selective approval.
* **Auto:** uses AI model review to reduce prompts on frequently used tools.

This per-tool configuration enables balancing security and user efficiency.

### Approval Patterns

Approval patterns define which commands require approval under HITL mode. Patterns use three match tiers:

* **Exact pattern match:** command matches a precise string or regex.
* **Relaxed fallback:** key words appear in order, allowing for flags or values in between.
* **Keyword matching:** command contains specified write verbs.

The default policy applies hard gates to destructive commands and soft gates to routine ones. Users see approval cards with options to approve once or always approve for soft gate actions, enabling saved approvals that prevent repeated prompts.

### Access Outside Workspace

Any action accessing paths outside the user's workspace but within their home folder always requires user approval, with options to save approval unless overridden by hard gates.

### Auto Approval

When a tool gate is set to Auto:

* Actions under hard gates always prompt the user.
* An AI model reviewer evaluates other actions and approves or escalates them.
* Escalations occur for destructive, irreversible, or ambiguous actions.
* Users can set chats to auto mode to leverage AI review, which improves efficiency by reducing prompts on safe actions.

Because AI decisions may vary, sensitive actions should remain under hard gates to ensure human review.

## How Actions Can Run Without User Prompt

Actions can execute without prompts in these ways:

* **Saved approvals:** users previously approved soft gate actions and saved the decision.
* **Model review:** AI reviewer approves actions not under hard gates, including commands not matching any pattern.
* **Tool gate mode allow:** tools set to Allow run actions without any approval.

However, hard gate actions or tools set to Deny always require user intervention or are blocked. Read-only commands are exempt from prompts to facilitate smooth routine workflows.

## Practical Benefits for ServiceNow Customers

* Granular control over which actions require approval ensures security without unnecessary interruptions.
* Balancing human oversight with AI review enhances operational efficiency, especially for frequent routine commands.
* Saved approvals reduce repetitive prompts for low-impact actions, improving user experience.
* Policy-driven enforcement ensures consistent governance and compliance with organizational rules.
* Flexible per-tool settings allow tailoring approval strictness based on risk profiles of different tools.
\</  
Tool gates, approval patterns, and auto approval determine whether each action in Arc runs, requires approval, or is blocked.

Three controls determine whether an action runs. Tool gates set how Arc handles each tool. Approval patterns identify commands that need approval. Auto approval lets an AI model reviewer approve some actions without prompting the user.
Figure 1. Action evaluation flow in Arc  
Every action passes through these controls in the following order:

1. Arc checks for an active policy. If no active policy is available, Arc refuses the action.

2. Arc applies the tool gate for the tool. Allow runs the action, Deny blocks it, the (Human in the loop) HITL sends it to approval patterns, and
   Auto sends it to auto approval.

3. For HITL, Arc compares the command against approval patterns. If no pattern matches, the action runs. If a hard gate pattern matches, Arc prompts the user. If a soft gate pattern matches, the action runs when a saved approval exists. Otherwise, Arc prompts the user.

4. For Auto, Arc prompts the user if the action is under a hard gate. Otherwise, the model reviewer approves the action or escalates it to the user.

5. When prompted, the user approves the action to run it or denies the action to block it.

6. When the action runs, sandbox, network, connector scope, and file type rules limit what the action can access.

{#action-approval-flow-cowork__ol_nnx_4g4_rkc}

## Gate strengths {#action-approval-flow-cowork__section_pgs_41m_5kc}

Some actions require a human decision before they run. Approval patterns, described later in this topic, identify these actions and assign each one a gate strength:
{#action-approval-flow-cowork__table_ecb_51m_5kc__entry__3}

| Gate | Behavior | Use |
|-|-|-|
| Hard gate | Requires a human decision each time the action runs. | Destructive actions, such as actions that delete data or make changes to external systems |
| Soft gate | Requires a human decision, which the user can save for later occurrences of the action. | Routine actions with low impact that users repeat |
[Table 1. Gate strengths]

{#action-approval-flow-cowork__table_ecb_51m_5kc}

## Tool gates {#action-approval-flow-cowork__tool-gates}

You set a mode on each tool. The modes and the tools each mode suits are described in the following table.  
{#action-approval-flow-cowork__table-tool-gate-modes__entry__3}

| Mode | Behavior | Use |
|-|-|-|
| Allow | Runs the action with no pattern matching and no prompt. | Tools with low impact, such as tools that only read data |
| Deny | Blocks the action with no option to approve it. | Tools that your organization doesn't permit |
| Human in the loop (HITL) | Checks the action against approval patterns. | Tools where specific commands require human review |
| Auto | Sends the action to auto approval. | Frequently used tools where repeated prompts slow users down |
[Table 2. Tool gate modes]

{#action-approval-flow-cowork__table-tool-gate-modes}

Because Auto is a gate mode, you control it per tool. You can reduce prompts on one frequently used tool while keeping stricter handling on others.

## Approval patterns {#action-approval-flow-cowork__approval-patterns}

Approval patterns identify the commands that need approval when a tool gate is set to HITL. The Default Policy includes predefined approval patterns organized by action category. Each pattern uses one of the
match tiers in the following table.  
{#action-approval-flow-cowork__table-match-tiers__entry__3}

| Tier | You configure | Matches when |
|-|-|-|
| Pattern matching, exact | A literal string or regular expression | The command matches the pattern as written. |
| Pattern matching, relaxed fallback | A pattern of several words | The words appear in the command in order, even with flags and values between them. |
| Keyword matching | A list of write verbs | The command contains one of the listed verbs. |
[Table 3. Approval pattern match tiers]

{#action-approval-flow-cowork__table-match-tiers}  
Note:  
Under HITL, a command that matches no approval pattern runs without a prompt. If you rely on keyword matching, a command that writes data with a verb that isn't in your list runs without approval.

Use exact matching when you know the precise command to control. Exact matching can miss variations of that command. Use relaxed matching to catch variations, but relaxed matching can match a word that appears where you didn't intend.
Use keyword matching to require approval for any command that writes data.

Arc applies a fixed guard on credential paths that you can't turn off.

Each pattern has a hard gate or a soft gate, as described in the gate strengths table.

The Default Policy applies hard gates to destructive operations and actions that delete data or make changes to
external systems, and soft gates to selected routine patterns with low impact.

When a pattern requires approval, Arc holds the action and displays an approval card to the user. The approval card offers Allow for every matched action and Always Allow only for soft gate patterns.

When a user selects Always Allow, Arc saves the decision and runs later actions of the same type without a prompt, such as the same command run on different files. An action of a different type prompts the user again. Users can view and revoke saved
approvals in SettingsApprovals.

## Access outside the workspace {#action-approval-flow-cowork__outside-workspace-access}

When an action reads or changes a path outside the user's workspace folders but inside the home folder, Arc always asks for approval. The approval card offers Always Allow unless the policy would apply a hard gate to the same action on its own, in which case the user approves it each time. Because
the decision comes from the policy, a change to your approval patterns changes this behavior too.

## Auto approval {#action-approval-flow-cowork__auto-approval}

When you set a tool gate to Auto, Arc prompts the user for any action under a hard gate. An AI model reviewer evaluates every other action, including actions that match no pattern, against the user's request.

The model reviewer escalates actions that are destructive, irreversible, or visible outside Arc. Some of these actions escalate every time. Others escalate unless the user's request clearly identifies what to act on. For example, "close INCXXXXXXX" identifies a specific record, but "clean up the incidents"
doesn't. An approved action proceeds. An escalated action appears to the user as a standard approval card.

The model reviewer is an AI model, and its decisions may be inaccurate or inappropriate. Because the reviewer evaluates context, its decisions can vary. To keep
a human review step for sensitive actions, place those actions under hard gates.

Users can
choose auto mode for each chat. A tool gate set to Auto works like HITL unless the auto approval capability is on and the user's chat is in auto mode. Steps that the reviewer approves show
Auto-mode allowed this.

## Ways an action runs without a prompt {#action-approval-flow-cowork__actions-without-prompt}

Under HITL or Auto, an action can run without a prompt in the ways described in the following table.  
{#action-approval-flow-cowork__table-actions-without-prompt__entry__3}

| Attribute | Saved approvals | Model review |
|-|-|-|
| Tool gate mode | HITL | Auto |
| Decision basis | A saved user decision for that type of action | An AI model's evaluation of the action in context |
| Set by | End user, one approval at a time | Administrator, by setting a tool gate to Auto |
| Scope | Soft gate patterns only | Actions not under a hard gate, including actions that match no pattern |
| Suited for | Repeated actions of a known, stable type | Actions that a static pattern can't classify in advance |
[Table 4. Ways an action runs without a prompt]

{#action-approval-flow-cowork__table-actions-without-prompt}

Neither method approves a hard gate action or runs an action on a tool set to Deny.

Arc also exempts read only commands from prompts, so routine work isn't interrupted.

*[\>]: and then


