Skip to content

Agent Types

Bonsai ships with 6 agent types, each specialized for a different role. One orchestrates; the rest execute.

Description: Architects the system, writes plans, reviews agent output — never writes application code

Best for: Project leads who need an agent that plans features, dispatches work to code agents, and reviews their output.

CategoryDefaults
Skillsplanning-template, issue-classification, dispatch, review-checklist
Workflowsissue-to-implementation, planning, code-review, session-logging
Protocolssession-start, security, scope-boundaries
Sensorssession-context, scope-guard-files, scope-guard-commands, dispatch-guard, subagent-stop-review
Routines(none by default — all 8 catalog routines are compatible)

Code agents execute plans created by the Tech Lead. Each has domain-specific skills and a focused scope.

Description: Executes backend plans — API, database, server-side logic

Best for: API development, database work, server-side business logic, and backend infrastructure.

CategoryDefaults
Skillscoding-standards, testing, database-conventions
Workflowsplan-execution, reporting, session-logging
Protocolssession-start, security, scope-boundaries
Sensorssession-context, scope-guard-files
Routines(none)

Description: Manages infrastructure-as-code, CI/CD pipelines, containers, and deployment automation

Best for: Terraform, Docker, Kubernetes, CI/CD pipelines, and deployment configuration.

CategoryDefaults
Skillscoding-standards, iac-conventions, container-standards
Workflowsplan-execution, reporting, session-logging, security-audit
Protocolssession-start, security, scope-boundaries
Sensorssession-context, scope-guard-files, iac-safety-guard
Routinesdependency-audit, infra-drift-check

Description: Executes frontend plans — UI components, state management, styling

Best for: UI component development, state management, CSS/styling, and frontend architecture.

CategoryDefaults
Skillscoding-standards, testing, design-guide
Workflowsplan-execution, reporting, session-logging
Protocolssession-start, security, scope-boundaries
Sensorssession-context, scope-guard-files
Routines(none)

Description: Implements full-stack features end-to-end — UI, API routes, database, auth, tests

Best for: Features that span the entire stack, when you want one agent handling frontend through database.

CategoryDefaults
Skillscoding-standards, testing, design-guide, database-conventions, api-design-standards, auth-patterns
Workflowsplan-execution, reporting, session-logging
Protocolssession-start, security, scope-boundaries
Sensorssession-context, scope-guard-files
Routines(none)

Description: Audits code for vulnerabilities, reviews auth patterns, scans dependencies, enforces security standards

Best for: Security audits, vulnerability scanning, dependency review, and auth pattern enforcement.

CategoryDefaults
Skillscoding-standards, testing, auth-patterns, review-checklist
Workflowscode-review, reporting, session-logging, security-audit
Protocolssession-start, security, scope-boundaries
Sensorssession-context, scope-guard-files, scope-guard-commands, api-security-check
Routinesvulnerability-scan, dependency-audit

If you need…Use this agent
Planning, orchestration, and code reviewTech Lead
API endpoints, database queries, server logicBackend
UI components, styling, state managementFrontend
End-to-end features across the full stackFull-Stack
Infrastructure, CI/CD, containersDevOps
Security audits, vulnerability scanningSecurity

Each agent is defined by an agent.yaml file and a core/ directory containing identity templates. See the agent.yaml Schema reference for the full format.