Agent Studio is where your organization's agents live: every task belongs to an agent, every agent has an identity and a dashboard, and task runs turn a request into finished work.
Agents and tasks
An agent is a configured helper — a name, a purpose, tools, and access. You address agents by handle in chat, and every task run belongs to the agent that executed it, so its history, dashboards, and outputs stay organized under it.
Task runs that produce documents
Task runs are multi-step: an agent can research, use tools, and iterate in a sandbox before it answers. Runs can produce real documents — presentations, PDFs, Word and Excel files, and images — via skill scripts in the run sandbox. Finished artifacts are attached to the task run, recorded in the run ledger, and included in notification emails, so a request like "assemble the weekly numbers into a deck" comes back as an actual file.
Dashboards
Each agent lands on a dashboard in Agent Studio, and dashboards are customizable — arrange the panels an agent's owner needs to see at a glance.
Connecting your data
- Store data — agents that work commerce questions can connect to your Shopify data through a read-only, scope-checked connection (see Store Agent).
- Code repositories — link one or more git repositories to an agent; task runs mount them alongside the agent, with credential checks at link time and pull-request support for Bitbucket Cloud. Repositories get their own tab on the agent.
- Apps — ClickUp, Microsoft Teams, and more. See Apps and channels.
Out-of-the-box agent templates
Rather than starting from a blank page, Agent Studio ships with ready-made agent templates — role-based AI colleagues with a persona, a desk of tables, and scheduled starter tasks that do real work on day one. Instantiate one, rename anything you like, and adapt it to your own data and tools.
| Template | Role | What it does |
|---|---|---|
| Penny 🛍️ | Merchandising Analyst | Keeps the catalog ready to sell — trend watch, competitor pricing, and inventory planning on a weekly loop. |
| Sterling 📦 | Operations Auditor | Watches every order and escalates what a careful ops lead would — anomalies, stuck fulfilments, silence after shipping. |
| Bell 📣 | Marketing Coordinator | Keeps promotions accurate and the calendar honest — campaign and promo checks with overlap and gap review. |
| Iris 💬 | Customer Insights Analyst | Tells the team what customers are actually asking — mines conversations for themes, sentiment, and product gaps. |
| Gauge 📊 | PMO Coordinator | Keeps every project honest — status, risks, and deadlines rolled up from your ClickUp workspace. |
| Nova 🔭 | GEO Specialist | Tracks how AI engines answer about your brand and who they cite instead — weekly citation waves across Perplexity, ChatGPT, and Gemini, with Semrush-enriched query sets. |
A few things worth knowing about how templates behave:
- Everything is editable. A template is configuration, not code: names, handles, charter text, schedules, and desk tables all adapt to your organization after instantiation.
- Starter tasks degrade gracefully. Tasks that need a specific connection (ClickUp for Gauge, Semrush for Nova) are born as drafts until that connection is set up — the rest of the agent works immediately.
- Measured, budgeted autonomy. Nova's probing follows the blind-prompt rule (AI engines are never told which brand is being measured) and runs under a monthly budget your organization controls.
New templates ship over time; if you need a role that is not on the list, contact us — custom agents and tools are how the catalog grows.