main*
📝agentful-parallel-agent-orchestration.md
📅March 14, 20253 min read📁agentful

Building Agentful: Parallel Agent Orchestration for Claude Code

#agentful#claude-code#ai-agents#mcp#typescript

Building Agentful: Parallel Agent Orchestration for Claude Code

The Problem

Claude Code is incredibly powerful, but when you're building a full-stack feature, a single agent becomes a bottleneck. The frontend agent doesn't know what the backend is doing. Tests get written after the fact. Code review happens at the end instead of continuously.

I wanted multiple agents working in parallel, each specialized, with shared context and inter-agent communication.

What Agentful Does

Agentful is a pre-configured development toolkit for Claude Code that orchestrates specialized agents working simultaneously in git worktrees:

Loading diagram...

Three-Tier Agent Architecture

Agentful uses a three-tier system:

  1. Core agents — Pre-built, always available (orchestrator, reviewer, fixer)
  2. Domain agents — Generated for your specific stack (frontend, backend, etc.)
  3. Ephemeral agents — Task-specific, created on-the-fly

Commands

CommandDescription
/agentful-initInteractive onboarding - 7 guided questions
/agentful-productCreate and analyze product specifications
/agentful-generateGenerate domain-specific agents for your stack
/agentful-startStart autonomous development
/agentful-statusView completion % and current work
/agentful-validateRun quality checks
/agentful-decideAnswer blocking decisions

Quality Gates

Every change passes through automated validation:

  • Types — TypeScript, Go, Rust type checking
  • Linting — ESLint, ruff, golangci-lint
  • Tests — Jest, Vitest, Pytest, Playwright
  • Coverage — Enforced thresholds
  • Security — Secret scanning, dependency audits
  • Dead code — Unused code detection

Pattern Learning

Enable agents to learn from every session:

claude mcp add agentful -- npx -y @itz4blitz/agentful-mcp-server

The reviewer stores error patterns, the fixer looks up known fixes, and the orchestrator stores successful implementation patterns — compounding knowledge across sessions.

Tech Stack Support

Auto-detects and adapts to your stack:

  • Languages: JavaScript, TypeScript, Python, Go, Rust, Java, C#, PHP, Ruby
  • Frontend: React, Vue, Angular, Svelte, Next.js, Astro, SolidJS
  • Backend: Express, Fastify, NestJS, Django, Flask, Spring Boot
  • Testing: Jest, Vitest, Playwright, Cypress, Pytest, JUnit

Quick Start

# 1. Install
npx @itz4blitz/agentful init
 
# 2. Start Claude Code
claude
 
# 3. Define product spec
/agentful-init
 
# 4. Start development
/agentful-start

Links