Skip to main content
Human in the Loop lets you require approval before your agent takes actions such as sending emails, creating repositories, or deleting records. Your approval settings determine which tool calls pause for review.

​
How it works

1

Agent pauses

When the agent tries to use a tool that requires approval, it pauses and shows an approval card with the tool and a readable summary of the proposed action.
2

You get notified

You receive an approval request via in-app notification, Slack DM, or directly in the chat thread.
3

You approve or reject

If approved, the agent continues from where it left off. If rejected, it acknowledges the rejection and adjusts its approach.
The agent never executes a gated tool call without your explicit go-ahead. Approval cards show readable resource names and key details, not every tool argument. Some details may be unavailable or hidden by your permissions. Review the task context as well as the card before approving.

​
Setting up approval controls

The most common approach. When you add a connector (like GitHub, Gmail, or Slack) to your agent, you can configure its Approval Settings to control when the agent needs permission.Open your agent’s configuration, find the connector in the Connectors list, and click the approval icon next to it.
Approval Settings dropdown showing Always allow, Ask each time, Ask for writes/deletes, Never allow, and Custom options
”Ask for writes/deletes” is a great default for most apps. It lets your agent gather information freely while keeping you in the loop for any changes.
Every connector’s tools are sorted into read-only and write when the connector is added, so Ask for writes/deletes works the same way on a custom or self-hosted connector as it does on Gmail or GitHub. If a tool is grouped in a way you disagree with, use Custom and set that one tool yourself.

​
What happens when a subagent pauses?

A subagent can pause for tool approval, an Ask Question card, or a step, time, or credit checkpoint. The request applies to that subagent. If the parent is waiting on that subagent, the parent pauses too until you answer. If a subagent has several pending tool actions, the parent conversation can show a card for each. Answer all required requests before the subagent can resume. Approving a tool does not answer a separate request for browser input or a checkpoint decision. Approval requests keep their assigned approver. Viewing the parent or subagent chat does not change who can approve them. When a background subagent pauses for tool approval, the request reaches you where you’re talking to the parent agent, and the parent keeps working meanwhile. In web chat, the approval card appears in the parent chat under Background subagents. In a Slack thread, use Approve or Reject on the request card. In iMessage or email, the notice reads Reply "approve" to approve, or review in Gumloop. Reply with just approve, or review it in Gumloop to decline. On Microsoft Teams, the request goes to your Gumloop inbox and a Slack DM.
1

Review the subagent request

Check which subagent is paused and whether it needs tool approval or has reached a checkpoint.
2

Choose the appropriate response

For a tool approval, choose Approve or Reject. For a checkpoint, choose Continue to resume the subagent, or Stop to end its run.
Checkpoint Continue and Stop apply to the subagent, not the whole parent task. Stopping returns a stopped result to the parent, not a successful completion. Continuing does not remove execution or concurrency limits.

​
Where you get notified

Approval requests appear in the chat and in your notifications. Slack-connected agents can also send requests in Slack.
A question or browser handoff asked in a web chat you’re watching does not also create an inbox notification.
The approval card appears directly in the conversation with the tool name, proposed action, key details, and Approve / Reject buttons. Use ⌘ + Enter to quickly approve.
Approval card in the web chat showing an email send request with To, Subject, and Body fields, plus Reject and Approve buttons
The card includes:
  • Tool name and icon at the top (for example, “Approve Send Email” with the Gmail icon)
  • Intent describing what the agent is trying to do
  • Display fields showing the key arguments (for example, To, Subject, Body)
  • Reject and Approve buttons
  • “Don’t ask again for this tool” checkbox to remember your decision for future calls

​
Example walkthroughs

Approving a GitHub tool call

1

Configure the app

Open your agent’s configuration, find GitHub in the Connectors list, and set the approval mode to Ask each time (or Ask for writes/deletes if you only want to approve changes).
2

Ask your agent to do something

For example: “Please create a PR on the docs repo to change the title of the agents skills doc page to just ‘Skills’.”
3

The agent pauses for approval

Before executing the GitHub API call, it shows you an approval card with the details:
Approval card for a GitHub tool call showing the intent, query, and Don't ask again checkbox with Reject and Approve buttons
4

Review and decide

Review the proposed action and its key details. If everything looks right, click Approve. If something is off, click Reject and optionally provide a reason.
5

The agent continues

After approval, the agent resumes work on the task.
1

Enable the App Rules Creation ability

Toggle it on in your agent’s Abilities section.
2

Ask your agent to create a rule

For example: “Create a human approval rule for any email that is sent to users without the gumloop.com domain.”
3

The agent builds the rule

It generates the appropriate CEL expression and shows you the rule details:
Agent creating a Gmail approval rule with a CEL condition that checks To/CC/BCC recipients for non-gumloop.com domains
4

Approve the rule

Once you approve, the rule is live. You can see it in the app’s config:
Gmail app configuration showing the newly created approval rule for external email recipients
5

The rule takes effect immediately

The rule requires approval for email to a non-gumloop.com address. Other approval settings and policies still apply, including to internal email.
Approval card showing the agent pausing to approve sending an email to an external recipient after the rule takes effect

​
What happens when you reject

When you reject a tool call:
  • The agent receives the rejection (and your optional reason).
  • It does not execute the tool.
  • It continues the conversation, often suggesting an alternative approach or asking for clarification.
Rejections are non-destructive. The agent simply adapts and tries a different path.

​
Approval modes at a glance

​
FAQ

The agent saves its state when it pauses for approval. Approving or rejecting resumes the request handling, but does not remove execution limits. Subagents can also pause at checkpoints; see What happens when a subagent pauses?.
Yes. If your agent is connected to Slack, approval buttons appear directly in the Slack thread and in a Slack DM. You can approve or reject right from Slack. You can also click “Open in Gumloop” to see the full context in the web app.
Checking this box before approving remembers your approval preference for that tool on the agent, not just the current request. Future calls use that preference, subject to applicable permissions and policies. For connector tools, you can change it in the connector’s Tool Management settings.For Manage Artifact, this applies to future artifact changes on the agent, not just the first file. File permissions and applicable policies still apply. There is currently no settings control to turn approval requests back on for Manage Artifact.
Approval settings are configured at the agent level and apply to all users interacting with that agent. For organization-wide policies that apply across all agents, use App Rules in your organization settings.
Approval Settings are simple toggles: always allow, ask each time, ask for writes, or custom per-tool. They apply based on the tool’s risk category (read vs. write).App Rules are conditional. They use CEL expressions to inspect the actual arguments of a tool call and decide whether to require approval. For example, “only require approval when the email recipient is outside my domain” is something only App Rules can do.Both can work together. Approval Settings act as the baseline, and App Rules add conditional overrides on top.
The Tasks page has an Approval Required filter that shows every pending approval across all your agents in one place. You can also use the notification bell in the top-right corner of the app for a quick glance.
The agent does not execute the rejected tool. It acknowledges the rejection, and if you provided a reason, it takes that into account. The agent then continues the conversation, often adjusting its approach or asking for clarification.
Yes, if you enable the App Rules Creation ability in your agent’s Abilities section. The agent can translate natural-language instructions into CEL conditions. The rule creation itself also requires your approval before it takes effect.
Yes. When an agent needs to run code in the sandbox (Python or shell), the approval card shows a preview of the code it wants to execute. You can review the exact script before approving. If the code makes MCP tool calls (for example, sending an email via Gmail), those nested calls are listed in the approval card as well.