Getting Started with Actors
Enable actor-based mode, select your virtual team, and pull the configuration into your workspace.
Enable actor-based mode
Actor-based mode is enabled per project in the Project Wizard or the Project Settings page.
New projects
In the Project Wizard, navigate to Step 4: Agent Setup. At the top of the step, two radio cards let you choose:
- Classic -- single agent with manual configuration
- Actor-Based -- virtual team with specialized actors
Select Actor-Based to enable the actor selection grid. The default actor ("Fullstack Development Architect") is automatically pre-selected.
Existing projects
Open the project in the Nexus dashboard and navigate to Settings > Agent Setup. Switch the agent mode from Classic to Actor-Based. The existing agent configuration is preserved -- actors are added on top of the runtime identity.
Switching modes does not delete your existing agent configuration. The project_agents entry (runtime identity) remains unchanged. Actor-based mode adds behavioral profiles on top of the runtime agent.
Select actors
In actor-based mode, the Agent Setup step shows an Actors tab with a selection grid:
- Each actor is shown as a card with avatar initials, display name, title, and domain badges
- Cards are arranged in a 2-column layout
- Click a card to select or deselect an actor
Selection rules
- Selecting any specialized actor automatically deselects the default actor
- The default actor shows a "fallback only" badge when disabled
- If all specialized actors are deselected, the default actor is automatically re-selected
- The default actor and specialized actors are mutually exclusive
Recommended approach
- Browse the available actors and read their role descriptions
- Select the actors that match your project's needs (e.g., a security architect for a security-sensitive project)
- Check the Skills tab to see which skills each actor recommends
- Adjust skill selection if needed -- remove skills you do not need or add skills not covered by any actor
Configure skills
After selecting actors, switch to the Skills tab. This tab shows all available skills with:
- Pre-selected checkboxes for skills recommended by your selected actors
- "Recommended" badges on skills that come from actor recommendations
- Per-actor skill count indicators (e.g., "Security Architect: 3 skills")
- The nexus-init skill is always included and cannot be removed
The skill selection is a union of all selected actors' recommendations. You can:
- Remove a recommended skill if you do not need it
- Add additional skills not recommended by any actor
- Keep the defaults for a curated experience
The same merge logic applies to plugins -- each actor's recommended_plugins are merged into the project plugin selection.
Pull configuration
After saving your project settings, pull the generated configuration into your workspace:
nexus pull
This generates or updates:
| File | Content |
|---|---|
.nexus/AGENTS.md | Agent identity with actor profiles, role definitions, and skill assignments |
.nexus/actors/*.md | Individual actor profile files (one per assigned actor) |
.nexus/actors.json | Machine-readable actor registry for the project |
opencode.json | OpenCode agent/subagent entries with model routing and permissions (if OpenCode adapter is enabled) |
The .nexus/ directory is automatically excluded from Git via .gitignore. Generated configuration files are workspace artifacts, not source-controlled files.
Config file generation
Both Classic and Actor-Based modes show a Config file generation section that controls:
- AGENTS.md -- auto-generated or manual upload override
- CLAUDE.md -- auto-generated or manual upload override
A warning dialog appears when you try to disable auto-generation, explaining that overriding may break:
- Confidentiality level enforcement
- Project directive materialization
- Plugin/MCP configurations
- Actor profile injection
Verify the setup
After pulling, verify the configuration is correct:
- Check that
.nexus/AGENTS.mdcontains your selected actors with their role definitions - Check that
.nexus/actors/contains individual profile files for each actor - If using OpenCode, verify
opencode.jsoncontains the expected agent entries - Start your agent and confirm it loads the actor context
# Check the generated files
ls -la .nexus/actors/
cat .nexus/actors.json | head -20
# Start your agent with actor context
opencode
The agent should identify itself using the primary actor's role definition and follow the behavioral guidelines from the actor profile.