The governance, verification & observability layer for the agentic SDLC.
Move at AI speed — and still trust, prove and audit what ships.
Website · Why Factory · Roadmap · Blog
This is the program/meta repository for the Factory family — cross-cutting plans, epics, and the shared PARR pipeline (Prepare · Act · Reflect · Review) that spans all the products. Work that touches a single product lives in that product's repo; work that spans the family lives here.
| Stage | Product | What it does |
|---|---|---|
| 🧭 Prepare / Plan + Review | PFactory | Governed planning grounded in live cloud/Backstage context, review gates with citations, human approval → governed GitHub issues |
| 🛠️ Act | AIFactory | Spec-first plan → code → QA in isolated worktrees; multi-provider; can delegate to other coding agents |
| 🧪 Reflect / Review | TFactory | Autonomous test generation + execution, graded on a 5-signal verdict (coverage delta, stability, mutation, lint, semantic relevance) |
| 🛰️ Review / Observe | CFactory | The control tower — one cockpit threading plan→code→test, with an advise-and-confirm copilot |
PFactory ──▶ AIFactory ──▶ TFactory … all observed & steered by … CFactory
(Plan) (Act) (Verify) (Cockpit)
AI code generation is commoditizing fast — and Factory deliberately does not compete there. We sit in the layer around the coding agents: govern → build → verify → observe.
The market is large and growing (AI code tools ≈ $8–10B in 2025–26 → $30–91B by 2033–35), but the value is migrating from writing code to trusting, governing and verifying it. The 2026 signal is stark: 84% of developers use AI coding tools, but only 29% trust the output; verification is the bottleneck; and the EU AI Act (high-risk rules from Aug 2, 2026) makes logging, human oversight and audit mandatory. That trust-and-governance gap is the layer Factory occupies.
Honest note: Factory is an early, open project — no market-share or revenue claims. We claim a position: the trust/governance/observability layer for the agentic SDLC.
- Complement, don't replace — wrap the coding agents you already use (Copilot, Cursor, Claude Code, Codex, Devin) rather than competing with them.
- Trust by construction — review gates + human approval (PFactory), a 5-signal test verdict (TFactory), and cross-pipeline observability (CFactory).
- Audit-ready — human-approval gates, HMAC-anchored audit logs and completion-event records: the evidence the EU AI Act asks for.
- Model-agnostic & future-proof — a multi-provider factory (Claude, OpenAI, Gemini, Ollama, vLLM, Codex, Copilot CLI), MCP interop, executor delegation, and BYO/air-gapped models. A new model or agent is a small adapter, not a rewrite.
→ Full positioning, real-life scenarios (solo dev → regulated enterprise), and the integration story: factory.freundcloud.com/why
docs/— the Jekyll site published at factory.freundcloud.com- Program planning & cross-cutting epics — see Issues and the Factory Program board
The family is managed as one program across five repos on a single
GitHub Project. Cross-cutting epics
live here in Factory; product-specific work lives in each product repo and is linked
up via cross-repo sub-issues. The current focus is the PARR spine (a shared
correlation key, a normalized completion-event schema, and a canonical port map) that
lets the four products cooperate — and lets CFactory observe them.