awslabs/aidlc-workflows

What it does

AI-DLC (AI-Driven Development Life Cycle) adds a structured delivery process on top of AI coding assistants. Instead of a single free-form prompt, it selects a workflow based on the request, asks for missing decisions, and pauses at human approval gates. The core is harness-neutral and ships integrations for Claude Code, Kiro CLI/IDE, Codex CLI, Cursor, opencode, and GitHub Copilot.

The methodology describes 5 phases and 33 stages, 14 agents (11 domain experts, 2 reviewers, 1 composer), and 11 workflow profiles covering features, bug fixes, infrastructure, security, and proofs of concept. It also keeps persistent state, team knowledge, learned rules, and an audit trail. If you want repeatable AI-assisted delivery with review checkpoints rather than ad-hoc chat, this is aimed at you.

Quick start

1. Install the latest stable release. On macOS, Linux, or WSL:

curl -fsSL https://github.com/awslabs/aidlc-workflows/releases/latest/download/install.sh | sh

On Windows, run PowerShell normally (not as administrator):

irm https://github.com/awslabs/aidlc-workflows/releases/latest/download/install.ps1 | iex

The installer provides the native aidlc command and the harness runtimes; Bun and Node.js are not required for this path. If you prefer to manage files yourself, download the aidlc-copy-runtime-X.Y.Z.tar.gz release and copy a complete runtime/<harness>/ directory into your project; that manual path requires Bun. Details are in Install and Lifecycle.

2. Configure a project from its root. For example, for Claude Code:

cd /path/to/your-project
aidlc config --harness claude
aidlc doctor

Replace claude with kiro, kiro-ide, codex, cursor, opencode, or copilot. Running aidlc config without --harness starts interactive setup. Then start a workflow inside the harness:

/aidlc Build a REST API for inventory management

Codex CLI uses $aidlc; in Kiro IDE, pick aidlc in the chat panel's agent picker. See the harness table and Getting Started for per-tool prerequisites and trust prompts.

Constraints and caveats

• Supported harness versions are specific: for example Codex CLI >= 0.145.0, opencode >= 1.18.29, and Kiro IDE >= 1.1.70. Check the harness-specific guide before installing.
• Model and provider configuration belong to the harness. Amazon Bedrock settings can be applied with aidlc config providers on supported surfaces, but Kiro needs no provider answer. The recommended model is Claude Opus 4.8.
• Human approval gates are built in, so the workflow intentionally pauses rather than completing autonomously.
• Kiro IDE asks whether to trust a folder; choosing trust allows that folder's .kiro hooks to run commands, so only trust projects you have reviewed.
• Generated AI output and costs should be reviewed before acting on them; AWS's responsible AI policy is linked from the project.
• The code is MIT-0 licensed.

HotRepo's take

The project entered the TypeScript daily list at #11 on 2026-11-05, with 5,694 stars and 1,025 forks and 25 stars gained that day. A strong launch signal for a tool announced in mid-October. The HN thread drew 29 points and 22 comments; one commenter recommended OpenSpec as an alternative that also handles brownfield projects, while another questioned whether elaborate skill/agent setups can underperform simply asking the model. The README's own framing — one core, many harnesses, with a deterministic engine — is the main differentiator here, and the approval-gate design is worth testing against your existing agent workflow.