The dashboard hub for building agents and the behavioral resources around them — domains, abilities, context blocks, MCP servers, and models.

The Agents area is the build hub for everything behavioral. It's where you create and edit agents and the resources they draw on.

Resources it contains and manages

Resource	What it manages
Agents	Core agent definitions (prompts, model assignment, heartbeat, scopes, attached abilities/domains/MCP)
Domains	User-activated modes that reconfigure an agent's behavior for a session
Abilities	Specialist sub-agents exposed as tools
Context blocks	Reusable prompt fragments and template variables
MCP servers	External tool servers registered for use by agents
Models	LLM configuration records (provider, model id, params) used for resolution

It's organized into tabs:

Tab	What you manage
Agents	Agent definitions — role, prompts, assignments
Domains	User-activated modes
Abilities	Specialist sub-agent tools
Context	Reusable prompt context blocks
MCP	External tool integrations
Models	LLM configuration records

The agent builder

Opening an agent gives you a focused editor, tab by tab:

Tab	What you set
Prompts	The system (soul) and developer prompts, plus runtime templates
Preview	See the composed prompt with context and variables resolved
Abilities	Assign specialist abilities the agent can invoke
Domains	Assign the user-activated modes available to the agent
MCP	Attach external tool servers
Heartbeat	Configure a recurring autonomous schedule
Scopes	Review the platform permissions the agent holds
Settings	Name, description, model assignment, and metadata

Changes you save here sync to your connected harness live — no restart needed.

For the concepts behind each of these, see Primitives. This page is about where to find and edit them in the dashboard.