Skip to main content

Why AIFactory

The problem​

AI can write code now. The models are good enough. What's missing isn't capability — it's trust and provenance, and that gap is widest exactly where the stakes are highest.

  • 96% of developers don't fully trust AI-generated code — yet only 48% verify it. (Sonar, 2026)
  • For 38% of teams, reviewing AI-written code now takes more effort than reviewing a human's.
  • ~74% of organizations can't produce security provenance for AI-generated code.

For a developer, that means babysitting a black box or cleaning up "vibe-code" you didn't plan. For a platform or security team at a regulated organization, it's worse: you can't send proprietary source to a third-party cloud, and you can't explain to an auditor where a given line of code came from. So the answer becomes "no" — and the productivity stays on the table.

What we're building instead​

AIFactory starts from a simple conviction: autonomy and governance are not opposites. You can have an agent that ships code and a trail you can defend.

That conviction shows up as five design commitments:

  1. Self-hosted, in your perimeter. It runs on your own infrastructure — Kubernetes via the Helm chart, or docker-compose on a laptop. Your code never has to leave your network.
  2. Spec-first. Every run begins with a written spec and acceptance criteria, not a vibe.
  3. Review-gated. You approve the plan before code is written and the diff before it merges. A QA agent checks the work against the spec.
  4. Isolated. Each task runs in its own git worktree; nothing touches your tree until you merge.
  5. Auditable. Every action is journaled in a hash-chained audit log; specs, plans, and QA reports live on disk and in version control. SOC2 / ISO evidence ships in the enterprise build.

And underneath all of it: no vendor lock-in. Route each phase to the model you choose — Claude, OpenAI, Gemini, Codex, or a local Ollama / OpenAI-compatible endpoint. You own your model bill, and you're never one provider's pricing change away from a rewrite.

Where we fit​

The autonomous-coding space is crowded at the top with well-funded, cloud-hosted, closed-source tools. AIFactory deliberately does not compete there. Its place is the intersection that those tools structurally can't occupy:

A full spec → plan → code → QA, review-gated autonomous pipeline, that you can self-host in your own infrastructure, with a deep, verifiable audit trail.

If you can use a cloud SaaS coding agent and you're happy merging unreviewed diffs, there are great tools for you — use them. AIFactory is for everyone who can't, or won't.

Open source, and open-core​

The core platform — the spec → plan → code → QA pipeline, the web UI, multi-provider routing, git worktree isolation, single-tenant self-hosting — is open source and free. That is the project, and it's the part most people will ever need.

A separate enterprise edition (multi-tenant isolation, SAML/SCIM, signed audit anchors with SOC2/ISO evidence export, policy packs, and support) exists for organizations that need it. The revenue from that is what keeps the open-source core healthy and maintained. We'd rather be honest about that model up front than pretend otherwise.

Our principles​

  • Candor over hype. We tell you what's production-grade and what's still beta. Regulated teams reward honesty, and so does everyone else.
  • Earn trust, don't claim it. The audit trail, the review gates, and the open code are there so you can verify what we say rather than take our word for it.
  • Demand-pulled, not feature-pushed. We don't add scope that no real user has asked for. If you have a use case we don't cover, open an issue first.

Next​