Create Your First Agent

Create an agent, understand what the agent page controls, and get it ready for real runs.

An agent is the runtime unit you operate in Salambo. It is the thing you expose through the API, monitor in the dashboard, and tune over time.

What You Set During Creation

When you create an agent, Salambo asks for:

Name: how the agent appears in the UI and API model list
Purpose: a short description of what the agent is responsible for

That is enough to create the record, but it is not enough to run successfully. After creation, you should immediately open the agent detail page and finish the runtime setup.

How To Create The Agent

Open Agents in the workspace navigation.
Choose Create agent.
Enter a clear, job-like name.
Add a short purpose so your team understands what this agent should handle.
Create the agent and open its detail page.

What To Configure Right Away

On the agent detail page, focus on these areas first:

Runtime: assign the snapshot the agent should boot from
Network access: decide whether the agent should use restricted mode or open internet
Environment variables: add any secrets or runtime values the agent needs
Telemetry: review OTEL export settings if your team uses external observability tooling

If you skip the snapshot assignment, the dashboard may flag the agent as needing attention because there is no runtime template attached yet.

How To Name Agents Well

Good agent names are specific and operational. Examples:

Lead Routing Operator
Support Inbox Triage
Client Report Generator
PI Research Assistant

Avoid vague names like Assistant, Runner, or Agent 1. The agent slug is also what appears as the model identifier in the v1 API, so naming matters beyond the UI.

Agent Statuses

You will normally see one of these states:

Active: available for new runs
Inactive: configured but not available for new runs
Archived: no longer part of the active operating surface

You can activate or deactivate an agent from its detail page. Deactivation stops new runs without deleting the agent configuration.

What To Do Next

After the agent exists:

attach a snapshot
add secrets and network rules
create an API key
start your first run

If you still need a snapshot, continue with Snapshots.

What You Set During Creation

When you create an agent, Salambo asks for:

Name: how the agent appears in the UI and API model list

Purpose: a short description of what the agent is responsible for

That is enough to create the record, but it is not enough to run successfully. After creation, you should immediately open the agent detail page and finish the runtime setup.

What To Configure Right Away

On the agent detail page, focus on these areas first:

Runtime: assign the snapshot the agent should boot from

Network access: decide whether the agent should use restricted mode or open internet

Environment variables: add any secrets or runtime values the agent needs

Telemetry: review OTEL export settings if your team uses external observability tooling

If you skip the snapshot assignment, the dashboard may flag the agent as needing attention because there is no runtime template attached yet.

How To Name Agents Well

Good agent names are specific and operational. Examples:

Lead Routing Operator

Support Inbox Triage

Client Report Generator

PI Research Assistant

Avoid vague names like Assistant, Runner, or Agent 1. The agent slug is also what appears as the model identifier in the v1 API, so naming matters beyond the UI.

Agent Statuses

You will normally see one of these states:

Active: available for new runs

Inactive: configured but not available for new runs

Archived: no longer part of the active operating surface

You can activate or deactivate an agent from its detail page. Deactivation stops new runs without deleting the agent configuration.