AI Agent Management

Settings

Manage your agent’s name and image, model and prompt, message limits, labels, and canned responses.

Open your agent in the dashboard and select Settings from its menu. The page has six sections:

SectionPurpose
GeneralAgent ID, name, image, credits limit message, and agent deletion
AIModel selection and temperature
PromptInstructions and prompt variables
SecurityMessage rate limits and public or private visibility
LabelsShared labels for organizing conversations
Canned responsesReusable messages for support operators

For widget appearance, initial messages, and channel connections, open Deploy. Startup forms and data collection are managed in Actions.

General

Agent details

The Agent Details card shows the agent’s ID and name:

  • Agent ID: The unique identifier for this agent. Use the copy button beside it when you need the ID for an integration or code.
  • Agent name: The name used to identify the agent in the dashboard. Edit it and select Save changes on the same card. The name is required and can contain up to 32 characters.

Agent image

Use the Agent image card to upload an image or remove the current one. The image appears in the dashboard and is saved when the upload finishes; this card has no separate save button.

Supported formats are PNG, JPEG, GIF, AVIF, APNG, and WebP, with a maximum file size of 5 MB.

Configure the widget’s display name and picture separately in widget content settings.

The English Muchat dashboard showing the settings menu, agent details, and agent image cards
General settings: section navigation, agent details, and agent image

Credits limit message

The Credits Limits Message card contains a text field and a Save changes button. You can save your preferred message here.

This card does not set a credit spending cap. The message is saved, but the current widget does not use it to display an out-of-credit message.

Delete agent

At the bottom of the page, Danger Zone contains Delete agent. Deleting the agent removes access to it and its connected channels.

Agent deletion cannot be undone

Confirm deletion only if you no longer need the agent. To pause responses temporarily on one channel, disable that channel in Deploy.

Select Delete, review the confirmation dialog, and select Delete again to proceed. Select Cancel to keep the agent.

AI

In AI, choose the response model and Temperature, then select Save changes.

OptionDescription
ModelChoose a model from the list. Access to some models depends on your subscription.
TemperatureSet a value between 0 and 1. Lower values, such as 0.2, produce more consistent responses; higher values, such as 1, produce more varied responses. A low value alone does not guarantee factual accuracy.

Saving on this page changes the main agent’s settings. To test models and instructions before applying them, use the Laboratory. See AI model for model descriptions and credit usage.

Prompt

Use Prompt to write the agent’s main instructions: its role, tone, and scope of responses. Edit the text and select Save changes at the top of the page.

See Instructions for guidance on writing instructions.

Prompt variables

Variables let you reuse a value throughout the prompt. To define or edit them:

Open prompt variables

Select Prompt Variables at the top of the page, then Add Variable.

Enter a key and value

Give each variable a nonempty, unique key and the value you want to use. Start the key with an English letter or underscore, followed only by English letters, digits, or underscores. For example, define response_language with the value fa.

Insert the variable into the prompt

Select Finish editing and save temporarily. Then click the variable above the editor to insert it into the text. The example appears as {{response_language}}; its value is substituted when the prompt is prepared for the model.

Save the changes

Select Save changes on the main prompt page to save both the text and the variables.

The response_language variable with the value fa and the finish editing button in Muchat’s prompt variables panel
Prompt variables: key, value, and finishing the edit

Finish editing and save temporarily only closes the variables panel and prepares the changes for saving. Variables are not saved to the main agent until you select Save changes.

If a channel has its own custom prompt, that prompt takes precedence over the main prompt for that channel.

Security

Rate limit

Use Rate limit to control how many messages the agent processes from each contact within a time window. Limits are counted separately for each contact on each channel, including the widget, dedicated page, Laboratory, API, and connected messaging channels.

  1. Enable Rate limit.
  2. Enter the maximum message count and the time window in seconds. Both values must be whole numbers of at least 1.
  3. Enter a Limit reached message for the limit warning. This text is displayed in the widget and dedicated page.
  4. Select Save changes at the top of the page.

For example, in the widget with 10 messages every 60 seconds, a contact who has sent 10 messages in the last 60 seconds must wait until the oldest message leaves that window before sending another. Starting a new conversation with the same contact identity does not reset the count.

For WhatsApp, Eitaa, Rubika, and Bale account connections, messages above the limit are kept in the inbox but do not receive automatic replies. The warning text is not sent to the contact on these connections.

To remove the limit, turn the switch off and save your changes.

Rate limit set to 10 messages every 60 seconds and the limit reached message in the English dashboard
Security: message count, time window, and limit reached message

Agent accessibility

The Agent Accessibility card sets Public or Private visibility. Private visibility restricts retrieval of the agent’s public information to users in the same organization.

Making an agent private does not disable published channels or replace visitor authentication. To stop access through the widget, dedicated page, or messaging platforms, disable the relevant channel in Deploy.

After changing visibility, select Save changes.

Labels

In Labels, create shared labels for organizing this agent’s conversations. Each label has a Name, Color, and optional Description.

Select Create label to add one; an empty list shows Create your first label. Enter a name, choose a preset or custom color, optionally add a description, and select Save.

Search the list to find a label. Use a row’s actions menu to edit or delete it. These labels are available when organizing conversations and configuring view and archive filters.

Before deleting a label

Deleting a label removes it from conversations, contacts, and messages. If a saved view uses the label in a filter, update that filter before deleting the label.

Canned responses

Canned responses are reusable operator messages stored with a Shortcode. Operators can insert them while replying in the inbox.

Create a canned response, find it with /, and insert it into a conversation

Create a canned response

Select Create canned response, or Create your first canned response if the list is empty.

Enter the short code and response text

Enter a unique short code, such as welcome, and the response text. The short code can contain up to 100 characters, and the text up to 20,000 characters. Then select Save.

Insert the response into a conversation

Type / in the inbox message editor, then search for and select a response. Its text is inserted into the editor. Review or edit it before sending.

You can use these variables in the response text:

VariableValue when inserted into a conversation
{{contact_name}}Contact’s name
{{contact_email}}Contact’s email
{{contact_phone}}Contact’s phone number
{{operator_name}}Operator’s name

To insert a variable in the response editor, type two opening braces and choose a variable from the list. If a value for one of the variables above is unavailable in that conversation, it is replaced with an empty string. Review the message before sending.

Use search and each row’s edit and delete buttons to manage existing responses.

MuChat documentation

On this page