Verified project record
awslabs/aidlc-workflows
aidlc-workflows turns AI coding assistants into structured, verifiable software-delivery workflows through a harness-neutral core with a deterministic engine, 5 phases and 33 stages, human approval gates, and a 105-event audit trail. It runs natively across seven coding harnesses including Claude Code, Kiro, Codex CLI, Cursor, opencode, and GitHub Copilot.
Project overview
A single harness-neutral core with per-harness configuration, 14 specialized agents, 11 workflow profiles, and 79 authoring and engine tools addresses the loss of context and structure that ad-hoc AI coding suffers as projects grow.
- Project type
- AI Agent · AI Coding · Workflow
- Use cases
- Coding & Development · Automation
- Deployment
- Refer to project documentation
- License
- MIT-0
Best for
- Developers, AI engineers, and teams who want their existing AI coding assistants to follow a structured, verifiable software-delivery workflow with human approval gates and an audit trail.
- Teams using multiple coding harnesses that want one harness-neutral methodology and deterministic engine across them.
Key capabilities
- Harness-neutral methodology and engine providing 5 phases and 33 stages from initialization through operation, driven by user requests in a coding agent.
- Runs natively in Claude Code, Kiro CLI, Kiro IDE, Codex CLI, Cursor, opencode, and GitHub Copilot with per-harness configuration.
- 11 domain experts, 2 reviewers, and an adaptive composer that AI-DLC uses within workflows.
- Profiles for features, bug fixes, infrastructure, security, proofs of concept, enterprise delivery, and other common work, including Classic and Express.
- Stops at approval gates before moving forward and asks for missing decisions, with source-bound review evidence.
- Records workflow events plus persistent state, team knowledge, and learned rules.
- Native aidlc command for configuring harnesses per project, running diagnostics (aidlc doctor), and managing providers and plugins.
- 79 aidlc-*.ts engine and authoring tools under core/tools for reshaping stages, agents, rules, sensors, and knowledge.
Limitations and risks
- Workflow execution requires one of the supported coding harnesses; the project itself is steering rules and an engine for those harnesses, not a standalone runtime.
- The workflows work best with capable reasoning models; the current recommended model is Claude Opus 4.8.
- Generative AI can make mistakes; users must review generated output and costs before acting on them.
Getting started
- Run the curl or PowerShell (irm) one-line installer for your OS, which adds the native aidlc command and all harness runtimes without requiring Bun or Node.js. Then cd to your project root, run aidlc config --harness <harness> and aidlc doctor for first-run diagnostics, and open the harness to invoke /aidlc with a work request.
Evidence and sources
- README: 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…
- README: The installer adds the native `aidlc` command and every harness runtime. Bun and Node.js are not required.
- README: From the project root, select the harness you use: ```bash cd /path/to/your-project aidlc config --harness claude aidlc doctor ```
- README: AI-DLC selects a workflow from the request, asks for missing decisions, and stops at approval gates before moving forward.
- README: Model-provider setup belongs to the harness. Shipped project configuration keeps the provider and model already selected by the user. `aidlc config providers` can apply Amazon Bed…
Search
Find projects, verify facts, compare options, and filter by your requirements
Try a searchA click only fills the search box; you stay in control
Project Details
0