ai-sdlc

An open-source orchestrator that drives AI coding agents through the full software development lifecycle — with quality gates, progressive autonomy, and codebase-aware context at every step.

The AI-SDLC Framework takes issues as input and routes them through a declared pipeline of stages, assigning AI agents and/or human reviewers at each stage, enforcing quality gates, and continuously learning which agents can be trusted with what.

The Problem

AI agents can build small greenfield projects, but software falls apart as it grows. Technical debt compounds, complexity overwhelms context windows, and developer velocity collapses:

• Productivity paradox: Experienced developers using AI tools are 19% slower on mature codebases, despite believing they are 20% faster (METR 2025)
• Quality decline: Refactoring dropped from 25% to 10% of changes; code churn rose from 5.5% to 7.9% (Gitclear 2024)
• Stability regression: Every 25% increase in AI adoption correlates with 7.2% drop in system stability (Google DORA)
• Trust gap: Only 3% of developers express high trust in AI output (Stack Overflow 2025)

The root cause isn't that AI agents write bad code. It's that nobody orchestrates how they work as the codebase grows.

How It Works

The orchestrator implements a continuous reconciliation loop:

1. WATCH    — Listen for triggers (issue assigned, CI failed, schedule)
2. ASSESS   — Analyze codebase complexity, score task complexity (1-10)
3. ROUTE    — Select strategy: fully-autonomous / AI-with-review / human-led
4. EXECUTE  — Invoke agent with context, constraints, and sandbox
5. VALIDATE — Run quality gates (tests, coverage, security, lint)
6. DELIVER  — Create PR with provenance, request review if required
7. LEARN    — Record outcome, update autonomy level, store episodic memory

Architecture