Agents

The Agents view lists the specialised assistants configured for this project. An agent is a named set of instructions, starter prompts and permissions that shapes how the assistant behaves for a particular kind of task -- a screenplay critic reads differently from a shot planner. For what the assistant is and how you talk to it, see The AI Assistant.

The list

Agents are listed by Name. Star the ones you reach for often, and the list keeps recently used agents to hand. Click an agent to open it for editing, or use the row menu to delete it. On an empty project the page offers Create your first AI agent; otherwise New Agent adds one.

Editing an agent

Opening an agent gives you:

Field	What it does
Name	What the agent is called in the list and in the assistant, e.g. "Screenplay Critic"
Description	A brief line to help you pick the right agent later
System Instructions	The prompt that defines this agent's behaviour and personality -- the substance of what makes it different
Suggested Prompts	Starter questions offered when you open a chat with this agent, e.g. "What should I do next"
Entity Access	What the agent is allowed to do with your project data -- see below

Save keeps the changes, Cancel discards them, and Delete removes the agent after a confirmation.

Entity access

Two checkboxes decide how much of the project an agent can touch:

• Entity Read Access -- the agent can read your entities: look up a character, list what properties a shot has, and link you to it.
• Entity Read-Write Access -- the agent can also change things: set properties, create entities, and delete them. This includes read access.

The two are wired so you cannot end up with an impossible combination. Granting write access implies read; taking read access away removes write with it. Give an agent write access only when you want it to change your project rather than talk about it.