awslabs/aidlc-workflows: AI-Driven Life Cycle (AI-DLC) adaptive workflow steering guidelines for AI coding brokers · GitHub

AI-DLC (AI-Driven Development Life Cycle) turns AI coding assistants into
structured, verifiable software-delivery workflows. One harness-neutral core
runs natively in Claude Code, Kiro CLI, Kiro IDE, Codex CLI, Cursor, opencode,
and GitHub Copilot.

The Quick Start beneath installs the newest steady AI-DLC launch.

macOS, Linux, or WSL:

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

Windows PowerShell:

irm https://github.com/awslabs/aidlc-workflows/releases/newest/obtain/set up.ps1 | iex

The installer units up every part AI-DLC wants: the native aidlc command and
each harness runtime. You don’t want to put in Bun or Node.js; a local
set up runs its hooks and instructions by means of aidlc and doesn’t name bun.
On Windows, it installs for the present account
and routinely registers the bin listing in User PATH. Run it from a traditional
PowerShell window; one opened with “Run as administrator” will get a warning and a
immediate, since putting in as administrator is much less protected. Open a brand new terminal
if one other session can’t discover aidlc. To skip each persistent and
current-process PATH adjustments, use
-NoModifyPath.
Windows uninstall removes solely the User PATH entry recorded as installer-owned.
On macOS, Linux, or WSL, observe the installer’s PATH instruction if wanted.

Prefer to handle the challenge information your self, or can’t set up a local
executable? The manual-copy path is the one place Bun is required: set up
Bun, obtain aidlc-copy-runtime-X.Y.Z.tar.gz from the
release, and duplicate
the entire runtime// listing into your challenge. This path does
not require the native aidlc command.

From the challenge root, choose the harness you employ:

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

Replace claude with kiro, kiro-ide, codex, cursor, opencode, or
copilot. Running aidlc config with out --harness begins the interactive
setup when a terminal is on the market. If you employ Kiro IDE’s personal terminal in a
challenge folder you haven’t trusted but, Kiro first asks whether or not you belief it.
Choose Trust Folder & Continue solely in your personal challenge or one you have got
checked, as a result of trusting lets the folder’s .kiro hooks run instructions in your
machine; in any other case select Cancel and assessment the folder first (see
First run).

Open your harness within the configured challenge and describe the work:

/aidlc Build a REST API for stock administration

Codex CLI makes use of $aidlc as a substitute of /aidlc. In Kiro IDE, first select aidlc
within the chat panel’s agent picker. AI-DLC selects a workflow from the request,
asks for lacking choices, and stops at approval gates earlier than transferring ahead.

For supplier setup, belief prompts, and harness-specific stipulations, use the
information within the desk beneath. The full walkthrough is in
Getting Started.

Model-provider setup belongs to the harness. Shipped challenge configuration
retains the supplier and mannequin already chosen by the person. aidlc config suppliers can apply Amazon Bedrock settings on supported challenge surfaces or
document guide setup for different harnesses. Kiro CLI and Kiro IDE want no supplier
reply as a result of mannequin entry comes with Kiro. On Kiro CLI 2.24.1 or later, use
kiro-ide, which pins Kiro’s v3 engine; kiro is the legacy row (Kiro CLI
says how an present kiro challenge switches). The methodology itself is
provider-independent.

AI-DLC works finest with succesful reasoning fashions. The present really useful mannequin
is Claude Opus 4.8. If Opus shouldn’t be out there to you, or you aren’t positive the place
to set the mannequin and energy in your harness, see
Choosing a Model and Effort.

Ad-hoc AI coding loses context as initiatives develop. AI-DLC retains necessities,
choices, implementation, assessments, and operational work related by means of one
audited lifecycle:

• 5 phases and 33 levels from initialization by means of operation
• 14 brokers: 11 area specialists, 2 reviewers, and an adaptive composer
• 11 workflow profiles for options, bug fixes, development projects, safety,
  proofs of idea, enterprise supply, and different frequent work
• Human approval gates and source-bound assessment proof
• 119-event audit path plus persistent state, group data, and realized guidelines
• The similar deterministic engine throughout each supported harness

Start with Workflow Profiles to check
Classic, Express, and the centered workflows. See the
AI-DLC Workflows 2.0 Specification
for the structure and methodology.

• core/ – hand-authored, harness-neutral methodology and engine
• core/instruments/ – 93 aidlc-*.ts engine and authoring instruments
• harness// – skinny, harness-specific manifests and integrations
• plugins// – non-compulsory AIDLC plugins
• scripts/ – packaging, binary, installer, and launch tooling
• assessments/ – smoke, unit, integration, and end-to-end assessments
• docs/ – person, harness-engineering, and developer documentation
• dist/ and dist-release/ – generated, ignored native outputs

Edit core/ or harness//, by no means generated dist* output.

Install dependencies and generate each harness:

bun set up --frozen-lockfile
bun scripts/package deal.ts

Useful instructions:

bun scripts/package deal.ts <identify>     # generate one harness
bun scripts/package deal.ts --check    # determinism guard
bun assessments/run-tests.ts --ci       # smoke, unit, and integration
bun assessments/run-tests.ts --release  # full launch acceptance

See the Contributing Guide for the entire
improvement workflow and Porting to a New Harness
so as to add one other runtime.

Run aidlc physician from the challenge root first. Common fixes:

See Troubleshooting for diagnostic and
restoration procedures.