Agent Identity

Give your agent a role in your business.

Define an agent’s instructions, model, and allowed tools. Give it a stable endpoint and, with email enabled, its own address so people can reach it where they work.

Agent profileExample configuration

Support assistantAcme company environment

Instructions
Find relevant cases and help the support team understand the customer’s issue.
Selected tools
Search recordsRead ticket details
Actions that change systems
Disabled
Email identity
acme-support@mail.victoros.ai
Allowed senders
People at @acme.example

Illustrative profile. You configure the role, tools, and access.

Define who it is and what it can do.

An agent belongs to a company and has a purpose you define. Its instructions, connected credentials, and selected tools shape how it responds and which actions it can take.

A name and a clear role

Set the agent’s name and instructions around a specific job. Explain what it should do, which context matters, and how it should respond.

A model for the task

Choose a supported provider and model, then configure available settings such as response length and reasoning. The configuration stays with the agent.

Tools you select

Choose the integration actions the agent can call and attach the credentials it needs. Access to one tool does not automatically give it access to every action in that integration.

An explicit action policy

Actions that change external systems require a separate setting, which is off by default. Set a turn limit to bound how long an agent can continue working on one answer.

A stable way to reach it

Each agent has a stable endpoint ID. Connect it to your application and keep that reference even if you rename the agent. Calls still require company authentication.

A record of its work

Review messages, responses, run status, and errors. Usage records show reported model consumption so you can understand how an agent is being used.

An email address of its own

Reach an agent from your inbox.

When email is enabled, a company admin can give an agent an address on the configured VictorHub mail domain. An allowed sender emails the agent and receives its answer in the same thread.

  1. 01

    Set up the address

    Assign one email address to the agent. Its address stays stable when you rename the agent.

  2. 02

    Choose who can write

    Allow specific sender addresses or entire domains. An empty allowed-sender list permits nobody.

  3. 03

    Follow the conversation

    The agent uses the email thread’s conversation history to answer. Review its messages and run outcomes in VictorHub.

A few practical questions.

Who does an agent belong to?

The company environment owns the agent. It stays with the company when the person who created it leaves, and company members can work on its configuration.

Can an agent update records or send messages?

Yes, when the relevant tools and credentials are configured and the setting for actions that change external systems is enabled. That setting is a standing permission, not a request for approval before each action.

Can anyone email my agent?

No. Email must be enabled for the deployment and configured for the agent. You choose the permitted sender addresses or domains; an empty list allows nobody.

Does agent email support attachments or custom domains?

The current email feature handles message text and sends plain-text replies. Attachment processing and customer-owned email domains are not currently supported.

What job would you give your agent?

Tell us about the role, the systems it needs, and how your team would reach it. We can show you how those pieces fit together in VictorHub.

Explore more of VictorHub

Workflow Builder

A business process often crosses several systems. Bring its triggers, decisions, and actions onto one canvas so you can see how information moves and what happens next.