System Commands
System commands are the built-in capabilities the agent may use while it runs inside this app. Enabling a command allows the agent to call it; the page is the per-app allow list.
Open Luna, then select Agent Settings > System Commands.
Before You Enable Commands
- Commands apply to the agent conversation of this app; when the app runtime mode is the message channel, the front end shows the message stream instead of the agent conversation
- Enabling a command is the same as granting that capability to the agent, so review the risk level of write and high-risk commands first
- Commands that touch resources, such as scripts, integration flows, or forms, only work while those resources exist and are published
Enable Commands For An Agent
- Filter by command group and search for a command or its description
- Read the row: command, description, status, and risk
- Review the risk before enabling a command, especially the write and high-risk levels
- Enable or disable commands one by one, or select several rows and use batch enable or batch disable
- Watch the counter: it shows how many commands are enabled out of the total
- Open a command to read its full description before enabling it
Verify An Enabled Command
- Open the agent conversation of this app and ask the agent to perform the operation that the command covers
- Confirm the agent can complete it, then disable the command and repeat to confirm it is refused
- Check the counter on the page: it shows how many commands are enabled out of the total
Troubleshooting
| Symptom | What to check |
|---|---|
| The agent still cannot do the operation | The command must be enabled for this app, the required resource must be published, and the app runtime mode must show the agent conversation |
| The enabled count does not change | Saving failed; enable again and read the message |
| A high-risk command is needed | Open the command details, read what it changes, and enable it alone so the change is traceable |
Risk Levels
| Risk | Meaning |
|---|---|
| Read | Only reads data |
| Session write | Writes to the current conversation session |
| Memory write | Writes to the agent memory |
| Write | Writes business data |
| High risk | Destructive or otherwise sensitive operations |
Command Groups
Commands are grouped by area, for example app installation, organization and members, resource management, message templates, agent context, documents, file resources, data processing, artifacts, presentations, roles and permissions, Soln Studio, analytics, automation, OpenLink, workspace apps, account center, audit log, approval, forms, and space.
Approval Mode And Background Operations
Use the approval-mode selector in the Luna composer before starting a run. Approval required is the default and retains confirmation for commands that require approval. No approval lets this run execute authorized commands without asking for operation confirmation. It still checks enabled commands, permissions, inputs, and business workflow approvals; missing or ambiguous information may require a question.
Form-schema deletion runs as a background operation. Luna displays an operation card with its status and result, and resumes querying the same operation after reconnecting. Closing Luna does not cancel accepted work. Use the card's cancel action while the operation is queued or retrying; running work cannot be cancelled. Enable system.operation.get and system.operation.cancel when the agent needs to query or cancel these operations through commands.