Understanding Agent Console

An agent takes on a defined piece of content work and carries it through to a result. You describe what you need, or upload the file it works from, and the agent asks for anything else it needs before it produces the output. The output is a working result, such as draft copy or a spreadsheet of translations, rather than a suggestion you have to act on yourself.

Agent Console opens on a dashboard that shows a tile for each agent available to you:

Each piece of work you do with an agent is a run. A run holds the whole conversation, and any files you uploaded or the agent produced.

Agent Console keeps every run in History, so finished work stays available after you close it. Agent Console names each run for you, and records the agent and model that produced it, and any playbook it followed. For an agent that creates something in Storyteq CMP, the run also records the ID of what it created. Because the conversation is kept, you can return to a run and continue it when you need changes, rather than starting again and re-explaining the task.

Every response also carries a thumbs up and thumbs down. Storyteq reviews this feedback to check that the agents are working as expected. To find runs, continue them, and give feedback, use Find, continue, and give feedback on agent runs.

A playbook is a reusable set of instructions for a recurring task. It records the purpose of the output, its structure and length, the inputs the agent should ask the user for, and the brand context the agent should follow, such as your tone of voice and preferred terms.

With a general AI tool, you describe your tone of voice, preferred terms, and structure in every prompt, and the result depends on how fully you remember to do it. A playbook captures that guidance once. When a run follows a playbook, the agent applies the same guidance every time, so copy written by different people in a team comes out consistent.

You choose a playbook when you send your first message to an agent. The Copywriter Agent uses playbooks. During private preview, Storyteq creates your organisation’s playbooks, and everyone with access to Agent Console can view them.

Agents produce on-brand output by drawing on material your organisation sets up, rather than starting from a blank prompt. Each agent gets that material from the place that suits its task:

Agent Storage holds shared files that agents work from across many runs. Today, only the Menu Copy Checker Agent uses it: it reads the pricing spreadsheets that a Manager uploads there. Keeping these files in one place means each person running the agent works from the same, current data.

Agent Storage is its own section of Agent Console, and everyone with access to Agent Console can view it. Managers upload new files and new versions of a file there, and archive or delete files that are no longer needed. Archived files don’t appear when you run an agent, so people can’t pick an out-of-date file by mistake. The first time a file is used in a run, you map its data to what the agent needs.

Access to Agent Console comes from a role your administrator assigns, alongside your other roles in Storyteq CMP. Without an Agent Console role, you can’t open Agent Console. It has two roles:

Most people who use the agents are Members. Managers look after the shared files the agents work from.

Agent Console is a separate application that uses your existing Storyteq CMP login, so you need a Storyteq CMP account to use it. Once you have access, Agent Console can appear in the application switcher in Content Portal and Adaptation Studio, alongside the other applications you use. You can also open it directly at https://app.storyteq.com/agent-console/<client-handle>;, where <client-handle> is your organisation’s handle: the same name that follows /portals/ in your Content Portal address.

Each role’s permissions are listed in Agent Console permissions. Control access to Agent Console covers how administrators assign the roles.