Agents
What it is
Section titled “What it is”The Agents feature lets you define, generate, and manage teams of AI agents within a program. Each agent has a name, role, model assignment, specializations, and an execution history. Agents can be created manually, generated via a wizard, or dispatched individually to specific tasks.
Why it matters
Section titled “Why it matters”Complex engagements need more than one AI perspective. Agent teams let you assign specialized roles (architect, developer, QA, analyst) with different models and prompt configurations, then dispatch them to appropriate tasks for parallel execution.
Key concepts
Section titled “Key concepts”- Agent table:
AgentTablelists all agents with name, role, model, status, success rate, and last activity - Agent drawer:
AgentDrawerslides in to show full agent details with profile, config, history, and version tabs - Profile tab:
AgentProfileTabshows agent identity, role, and specializations - Config tab:
AgentConfigTabconfigures model, prompt parameters, and behavior - History tab:
AgentHistoryTabshows past execution records - Versions tab:
AgentVersionsTabtracks prompt and configuration changes over time - Agent avatar:
AgentAvatargenerates a unique visual identity per agent based onavatarSeed - Status badge:
AgentStatusBadgeshowsactive,idle,executing,archived, etc. - Create agent modal:
CreateAgentModalfor manual agent creation - Generate wizard:
GenerateWizardauto-generates a team of agents with three steps:WizardStepScope— define scope (sprint, workstream, or custom task selection)WizardStepRoster— AI generates a recommended team rosterWizardStepConfirm— review and confirm the generated team
- Dispatch modal:
DispatchAgentModalassigns a specific agent to a task for execution - Orchestration access: Launch the orchestration wizard directly from the agents page
How to use it
Section titled “How to use it”- Navigate to [Program] > Agents.
- View the agent table showing all agents, their roles, and performance metrics.
- Click Generate Team to launch the wizard — select scope, review the AI-recommended roster, and confirm.
- Click an agent row to open the drawer with profile, config, and history.
- Click Dispatch on an agent to assign it to a specific task.
- Click New Orchestration to launch a multi-agent run from the agents page.
- Archive agents that are no longer needed.
Data model
Section titled “Data model”This feature uses the following tables:
agentTeam/agents— Agent records with name, role, model, status, specializations, avatarSeedagentTeam/executions— Execution records per agent with input/output summaries and statusagentExecutions— Global execution log across all agents