Agent settings reference

This reference explains each setting that shapes how an agent works. The same settings apply to built-in and custom agents. On a built-in agent you change them in a custom version; on a custom agent you set them when you create it and edit its versions afterwards.

To find the settings, go to Automation > Agents, select an agent and open the Versions tab.

Identity

Setting	Applies to	What it does
Agent icon	Custom agents	The avatar shown wherever the agent appears.
Agent name	Custom agents	The name people see when choosing or running the agent. Required.

You set a custom agent's icon and name when you create it. Built-in agents keep their Helpin names, such as Atlas or Quill.

Behavior

Setting	What it does
Instructions (System prompt)	Tells the agent how to behave, what good output looks like and when to ask for help. Required for custom agents.
Skills (Available skills)	Reusable guides the agent can find and read during a run. If a skill needs certain tools, Helpin adds them automatically.

Write instructions as you would brief a new teammate: the goal, the steps you expect, the format of the result and anything the agent must not do.

Tools

Tools (shown as Allowed tools on existing agents) are the actions and data sources the agent may use, such as searching documents, listing tasks or searching the web. The tool picker groups them by category, for example Docs, PM / Tasks, CRM, Support and Web Search. Tools from external MCP servers your workspace has connected also appear here.

An agent can only use tools that are on this list. A tool added by a skill can't be removed while that skill is attached; remove the skill first. For more about tools and skills, see Browse the tool catalog and Skills and the skill catalog.

Working areas

Working areas control where the agent appears as an option you can start and what it can be started on:

Working area	The agent can be started on
Tasks, Epics, Sprints, Objectives	Project work items
CRM deals, CRM contacts, CRM companies	CRM records
Docs	Documents
Support	Support conversations
Workspace	The workspace as a whole, with no single record
Code repo	A connected code repository

Choose at least one. Pick the narrowest set that fits the job.

Access

Who can use this agent (custom agents only, set when you create the agent) controls who can see the agent in pickers and start runs:

All teams — anyone in the workspace.
Specific teams — only members of the teams you select, plus workspace owners and admins.

Access decides who can start the agent. Its tools and skills decide what it can do once it runs. See Who can use and manage agents.

Run mode

Run mode	Behavior
Autonomous	Runs end to end in the background.
Interactive	Can ask follow-up questions or request approval during the run.

The agent owns its default run mode. See Run modes: autonomous and interactive.

AI profile

The AI profile chooses which configured AI models the agent uses. Only profiles shared with the workspace are listed. If you don't choose one, the agent uses the Workspace default. The profile appears as Model size on the Agents page. Runs that have already been accepted keep the profile they started with.

Was this article helpful?