aws-solutions/konductor

---

Why Konductor?

Most AI coding assistants are one agent in one conversation. You end up as the project manager: writing the requirements, asking for a design, pasting it back for implementation, then remembering to ask for tests.

Konductor installs a team that already knows how to work together. Ask the konductor orchestrator for something in plain language. It breaks the request down, sends each piece to the right specialist, checks what comes back, and sends it back for fixes when it falls short.

• 🎼 One request, the whole lifecycle. Requirements, design, code, tests and docs, without hand-carrying artifacts between chats.
• ✅ Every deliverable gets checked. Work passes a maker-checker gate, and failed checks go back for up to two fix cycles.
• 🧠 Expertise built in. 82 skills and 19 step-by-step workflows (SOPs) cover threat models, IAM policies, DynamoDB design, code review, E2E testing and more.
• 🪶 No infrastructure. Agents are configuration (prompts, skills, SOPs) that your Kiro or Claude Code runtime runs directly. No servers, no extra accounts.

flowchart LR
    You(["💬 You"]) --> K["🎼 konductororchestrator"]
    K --> PM["📋 k-product-managerrequirements"]
    K --> AR["🏛️ k-architectdesign"]
    K --> DEV["🛠️ k-developercode"]
    K --> QA["🧪 k-quality-assurancetests"]
    K --> MORE["🔎 researcher · 📊 tpm🌐 browser · 🖼️ media"]
    PM & AR & DEV & QA & MORE -.->|"checked before handoff"| K
    classDef orch fill:#7C5CFF,color:#ffffff,stroke:#7C5CFF
    classDef agent fill:#0B0A12,color:#ffffff,stroke:#3DDC97
    class K orch
    class PM,AR,DEV,QA,MORE agent