> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sequentum.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents

> Reusable web automations that browse sites, extract records, and run on demand or on a schedule.

A Sequentum **Agent** is a reusable web automation that visits a site, follows instructions, and returns structured records. Agents can run manually, from the API, through MCP, or on a [Schedule](/concepts/schedules).

Most teams create Agents with [Agent Builder](/agent-builder/overview). Use [the Editor](/agents/overview) when you need full control over commands, scripts, or migration from Sequentum Enterprise.

## Two ways to author an Agent

<CardGroup cols={2}>
  <Card title="Agent Builder" icon="wand-magic-sparkles" href="/agent-builder/overview">
    Default path. Describe the site and fields in plain language; Sequentum plans, builds, tests, and hands you a runnable Agent.
  </Card>

  <Card title="The Editor" icon="code" href="/agents/overview">
    Advanced path. Hand-author the command tree for complex navigation, custom scripts, or niche commands.
  </Card>
</CardGroup>

## What an Agent contains

Every Agent combines:

* **Commands** — the navigation and extraction instructions: open pages, click, fill forms, paginate, parse fields, and export data.
* **Output schema** — the fields and record shape the Agent emits.
* **Runtime settings** — optional browser, proxy, anti-blocking, and execution properties.

Agents live inside a [Space](/concepts/spaces). They are versioned, so each save creates a new version you can restore later.

## Lifecycle

<Steps>
  <Step title="Build">
    A new Agent starts as a draft while it is being built or edited.
  </Step>

  <Step title="Save">
    Publishing the draft creates the first saved version. The Agent receives an integer `agentId` and `versionNumber` `1`.
  </Step>

  <Step title="Create versions">
    Each later save creates a new `versionNumber`. The highest-numbered version is active unless you restore an earlier one.
  </Step>
</Steps>

Archive an Agent to hide it without deleting it. Unarchive it when you need to run or edit it again.

Run lifecycle is separate. Starting an Agent creates a [Run](/concepts/runs) with its own `runId`, status, outputs, and diagnostics.

## Run an Agent

Once an Agent is saved, you can:

* Start it manually in Control Center.
* Call [`POST /agent/{agentId}/start`](/api-reference/agents/start-run) from your application.
* Attach a [Schedule](/concepts/schedules) for recurring runs.
* Expose it to LLM workflows through [MCP](/mcp/overview).

## Identifiers

These are the IDs you'll see most often when working with Agents:

<CardGroup cols={2}>
  <Card title="agentId" icon="fingerprint">
    Unique identifier for the Agent. It appears in Agent detail screens, run-start responses, and Agent API paths.
  </Card>

  <Card title="versionNumber" icon="code-branch">
    Saved version of an Agent. It increments on each save and is used when restoring an earlier version.
  </Card>

  <Card title="runId" icon="circle-play">
    Identifier for a single execution of an Agent. Each run has its own status, outputs, diagnostics, and lifecycle.
  </Card>
</CardGroup>

## Related API references

<CardGroup cols={2}>
  <Card title="Start a run" icon="play" href="/api-reference/agents/start-run">
    Trigger an Agent and optionally return records synchronously.
  </Card>

  <Card title="List agents" icon="robot" href="/api-reference/agents/list">
    Find the Agents available in a Space.
  </Card>

  <Card title="Restore a version" icon="rotate-left" href="/api-reference/agents/restore-version">
    Roll an Agent back to an earlier saved version.
  </Card>

  <Card title="Run lifecycle" icon="circle-play" href="/concepts/runs">
    Understand run status, outputs, stopping, and failures.
  </Card>
</CardGroup>