> ## Documentation Index > Fetch the complete documentation index at: https://docs.getoneprofile.ai/llms.txt > Use this file to discover all available pages before exploring further. # AI agent columns > Add a column that runs a prompt on every row, reads your other columns, and writes back one or more fields. An AI agent column runs a prompt against each row and writes the result into one or more fields. Use it where an enrichment or a function cannot help: research a company, classify a lead, or say why it fits. The agent column editor docked beside a table, showing an Effort card with a web research toggle, a Prompt card whose text holds a column reference pill, and an Outputs card listing two named fields. ## Add an agent column Click **Tools** in the toolbar, then the **Agent** tab. Click **Use AI** at the top of **Your agents** to write a new prompt from scratch. The editor opens on an **AI Assistant** card asking what this column should find or write. Describe what you want and click **Generate**, or click **Set up manually** to fill in the fields yourself. The **Effort** card sets which model the column runs on, and each option shows its cost per row. Turn on **Web research** to let the agent search the web, and **Use business context** to add your company profile to every run. In the **Prompt** card, type `/` to reference another column's value, for example "Write a one-line summary of /Company". **Add column** under the prompt inserts a reference without typing. In the **Outputs** card, pick **Fields** to list what comes back, each with its own name and type. Each field becomes its own column once the agent runs. Click **Preview on 3 rows** to run the column on the table's first three rows for free. Then use the split button for **Create & run 10 rows**, **Create & run 50 rows**, or **Create and don't run**. ## Build an agent you can reuse The **Agents** page is where you build agents to deploy on any table. It is one page with three sections: **Build an agent**, **Or start with a template** and **Browse agents**. Describe the agent under **Build an agent** and click **Generate**, or click **Set up manually** to start from empty fields. **Built-in templates** opens every agent Oneprofile ships, plus the ones your team has saved. ## Good to know **Your agents** comes first: **Use AI**, the blank prompt you write on the spot, then any agent you have saved. **Built-in agents** sits under it, ready to run without writing a prompt. Arrow keys move through the picker, and the right arrow drills into a column that holds several fields or list items. Click a pill to mark that reference required, so a row is skipped when the value is missing. Press Backspace to remove one. **Compare effort levels** in the **Effort** card opens the full trade-off between the models. The **Advanced** disclosure below it sets a **Starting URL column** the research begins from, and can keep the agent on that domain. Click **Add output** for another field, and **Add description** to explain one to the model. A description is worth writing when the field name alone could be read two ways. **JSON Schema** in the **Outputs** card replaces the field list with a raw schema you write yourself. **Preview on 3 rows** is free, within a daily allowance for your workspace. When the allowance runs out, the preview tells you so instead of running. A column deployed from a saved agent stays linked to it. Its banner names the agent and tells you to open the agent to change the prompt, effort, output or tools. The **Quick set-up** tab stays editable, so you can still remap which table columns feed the agent's inputs. Click **Stop using this agent** on the banner to keep a copy of its settings on the column instead. Preview stays available on a linked column. Previewing spends nothing and changes nothing. ## Next steps Fill a column from the catalog instead of a prompt. Control what runs, when, and over which rows.