Orchestrate Multi-Agent Code Analysis

Coordinates multiple specialized agents through Claude Code's native Agent Tool, keeping all orchestration within Claude's control rather than relying on external orchestration scripts.

Good for complex tasks requiring multiple expertise domains, code analysis from security/performance/quality perspectives simultaneously, comprehensive reviews spanning architecture/security/testing, or feature implementation needing backend, frontend, and database work together. Not for simple single-domain tasks, quick fixes, or small changes where one agent suffices.

Native invocation patterns include single-agent calls ("Use the security-auditor agent to review authentication"), sequential chains explicitly naming each agent step in order, context-passing invocations where one agent's findings feed the next, and resuming a previous agent by ID with additional requirements.

Three orchestration patterns are defined: Comprehensive Analysis (explorer-agent maps the codebase, then security-auditor, backend-specialist, frontend-specialist, and test-engineer each analyze their domain, followed by synthesis); Feature Review (identify affected domains, invoke the relevant domain agents, have test-engineer verify changes, then synthesize recommendations); and Security Audit (security-auditor performs configuration/code review, penetration-tester runs active vulnerability testing, then synthesis with prioritized remediation).

Lists seventeen available specialized agents with their expertise and trigger phrases: orchestrator (coordination), security-auditor (security/auth/vulnerabilities), penetration-tester (pentest/red team/exploit), backend-specialist (API/server/Node.js/Express), frontend-specialist (React/UI/Next.js), test-engineer (tests/coverage/TDD), devops-engineer (deploy/CI/CD/infrastructure), database-architect (schema/Prisma/migrations), mobile-developer (React Native/Flutter), api-designer (REST/GraphQL/OpenAPI), debugger (bug/error), explorer-agent (explore/map/structure), documentation-writer, performance-optimizer (slow/optimize/profiling), project-planner (plan/roadmap), seo-specialist, and game-developer (Unity/Godot/Phaser). Notes Claude Code's built-in agents work alongside custom ones: Explore (Haiku, fast read-only search), Plan (Sonnet, plan-mode research), and General-purpose (Sonnet, complex multi-step modifications) - use Explore for quick searches and custom agents for domain expertise.

The synthesis protocol requires a single unified report after all agents complete: a task summary, an agent-contributions table mapping each agent to its finding, consolidated recommendations tiered by severity (critical/important/nice-to-have), and a concrete action item checklist.

Best practices recommend following a logical order (discovery to analysis to implementation to testing), sharing context between agents, producing one unified synthesis rather than separate outputs, and always including test-engineer to verify code modifications. Key benefits include single-session shared context, AI-controlled autonomous orchestration, native integration with built-in Explore/Plan agents, resume support for continuing prior agent work, and context passing between agents.

Works natively within Claude Code's Agent Tool, coordinating both custom specialized agents and Claude Code's built-in Explore/Plan/General-purpose agents in a single session.

Engineers using Claude Code who need multi-domain analysis or review coordinated natively across specialized agents with a single synthesized output, rather than running isolated single-agent tasks or external orchestration scripts.

First, use the explorer-agent to discover project structure.
Then, use the backend-specialist to review API endpoints.