Agent Editor
Create and edit AI agents. Define system prompts that shape Claude's personality, behavior, and capabilities.
On this page
The agent editor lets you create new agents or modify existing ones. Agents define Claude's behavior through system prompts — the foundational instructions that shape every response.
Form Fields
Name
A descriptive name for the agent that reflects its role or persona. Good names make it clear what the agent specializes in — for example, "Senior Code Reviewer" or "Marketing Copywriter".
Description
A brief explanation of the agent's purpose and capabilities. This appears alongside the agent name in the agents table and helps you quickly identify which agent to use for a given task.
System Prompt
The system prompt is the core of an agent. It defines who Claude is in this context, how it should behave, and what rules it should follow. The system prompt is sent to Claude at the start of every conversation using this agent.
Writing Effective System Prompts
- Define the role — Tell Claude who it is and what expertise it has
- Set the tone — Specify the communication style (formal, casual, technical, friendly)
- Establish boundaries — Be clear about what the agent should and should not do
- Provide context — Include relevant background knowledge or domain information
- Structure the output — Describe how responses should be formatted
- Include priorities — State what matters most (accuracy, speed, creativity, thoroughness)
Plugin Assignment
Optionally assign the agent to one or more plugins. This bundles the agent with related skills and MCP servers. When users install the plugin, they get the agent along with everything else in the package.
Saving Changes
Save your changes to update the agent immediately in your workspace. If the agent belongs to a published plugin, the changes will be included in your next marketplace publish.