Agents
AgentLoop coordinates multiple specialized AI agents that work together to complete software development tasks. Each agent has specific capabilities and responsibilities.
Agent Types
Orchestrator Agent
The orchestrator agent manages overall workflow, delegates tasks, and ensures smooth handoffs between agents.
Responsibilities:
- Coordinate task execution
- Manage agent delegation
- Handle workflow transitions
- Monitor progress and recovery
Product Manager Agent
The product manager agent breaks down project descriptions into structured, actionable work.
Responsibilities:
- Task breakdown into AGILE format
- User story creation
- Sprint planning
- Requirements gathering
Example usage:
> Break down "user authentication" into tasks and add them to the board> Create user stories for the checkout flowEngineer Agent
The engineer agent handles code implementation with proper git workflows.
Responsibilities:
- Code implementation
- Git operations (commit, branch, merge)
- Pull request creation
- Code modifications
Example usage:
> Have the engineer agent implement the login API endpoint> Ask the engineer to fix the bug in the payment moduleQA Tester Agent
The QA agent validates completed work and ensures quality standards.
Responsibilities:
- Test execution
- Issue identification
- Validation of implementations
- Bug reporting
Example usage:
> Ask the QA agent to test the new authentication flow> Have QA validate the API changes before marking them doneCode Analyzer Agent
The code analyzer performs deep analysis of the codebase for understanding and planning.
Responsibilities:
- Architecture analysis
- Dependency mapping
- Code pattern identification
- Impact assessment
Example usage:
> Let the code analyzer look at the payment module architecture> Analyze the dependencies in the auth systemDelegating to Agents
Interactive Mode
In interactive mode, delegate to agents using natural language:
> Have the engineer agent implement the API endpoint> Ask the QA agent to test the authentication flow> Let the code analyzer examine the module structureAgent Monitoring
View real-time agent status:
> /orchestrator agentsThis displays:
- Which agents are active
- What task each is working on
- Current progress
- Resource utilization
Parallel Agent Execution
Multiple agents can work simultaneously on independent tasks:
[orchestrator]
max_parallel_agents = 4With parallelism enabled:
- Independent tasks execute concurrently
- Each agent can work in its own git worktree
- Optional container sandboxing for isolation
See Git Worktrees and Container Sandboxing for isolation options.
Agent Configuration
Agents can be customized through the plugin system:
Custom agent templates - Create agents with specific behaviors using markdown templates
Behavior trees - Configure decision logic for complex workflows
MCP integration - Extend agents with external tools via Model Context Protocol
Agent Workflow
┌────────────────┐ ┌────────────────┐ ┌────────────────┐
│ Product │────▶│ Engineer │────▶│ QA │
│ Manager │ │ Agent(s) │ │ Tester │
│ │ │ │ │ │
│ • Break down │ │ • Implement │ │ • Test │
│ • Plan sprints │ │ • Commit │ │ • Validate │
│ • Create tasks │ │ • Create PRs │ │ • Report │
└────────────────┘ └────────────────┘ └────────────────┘
│ │ │
│ ▼ │
│ ┌────────────────┐ │
│ │ Code Analyzer │ │
│ │ │ │
│ │ • Analyze │ │
│ │ • Map deps │ │
│ │ • Assess │ │
│ └────────────────┘ │
│ │ │
└──────────────────────┼──────────────────────┘
▼
┌────────────────────┐
│ Orchestrator │
│ Coordination │
└────────────────────┘Quality Gates
The orchestrator enforces quality gates between agent handoffs:
- PM → Engineer - Tasks must have clear requirements
- Engineer → QA - Code changes must be complete
- QA → Done - Tests must pass validation
Quality gates are enforced automatically by the orchestrator’s behavior tree engine.