Production AI Institute · PSF v1.1 open standard
AI Right-To-KnowAI Data Use IndexCheck My AI ToolsPolicy Change WatchAgent ReadinessPublic BenchmarkContactGlobal standard · Worldwide
InsightsOpen source

Why We Open-Sourced WorkflowOS

Standards earn trust when practitioners can inspect the reference implementation — not just read the PDF. WorkflowOS is our answer: a free PSF workflow designer, MIT-licensed on GitHub.

Production AI Institute · 8 min read · Updated June 2026

The Production Safety Framework (PSF) defines what safe production AI deployment looks like across eight domains. That text is published openly and free to cite. What was missing was a working artifact — something teams could open, model a real workflow in, stress-test against PSF controls, and export as evidence.

WorkflowOS (PSF Workflow Studio) fills that gap. In June 2026 we released it under the MIT licence at github.com/troygamble/workflowOS. Use the hosted edition at /studio, or fork and self-host for client engagements.

View on GitHubOpen hosted Studio →

Why open source, not a paid product

Three reasons drove the decision:

  1. Credibility. An standards body that hides its reference tooling behind a paywall invites scepticism. Open source lets auditors, procurement teams, and practitioners verify that PSF language maps to real controls.
  2. MSP and integrator value. Certified AI Integrators deliver PSF-aligned work for clients. They need a tool they can self-host, rebrand exports from, and ship without licence friction. MIT removes that friction.
  3. Acquisition without lock-in. WorkflowOS is the front door to the PSF flywheel: use the tool → see gaps → study → certify. Revenue stays in credentials and organisational programmes, not in renting a canvas.

What you get

Visual workflow canvas
Six typed node kinds — triggers, skills, human gates, conditions, integrations — with contracts and validation notes.
PSF 8-domain analysis
Score workflows against every PSF domain before you ship. Gaps become visible on the canvas, not in a post-mortem.
Simulate and heal
Run deterministic simulations, use AI heal to repair broken graphs, and generate executive briefs from the same model.
Exportable evidence
YAML packages, client proposals, flow diagrams, and portfolio artefacts suitable for CPAP review.

Hosted vs self-hosted

Hosted Studio at productionai.institute/studio requires a free sign-in. Non-AI features work immediately. AI features use bring-your-own-key OpenAI credentials — add your key once in the Studio ⋯ menu. We do not supply a shared institute API key in production.

Self-hosted clones the public repo only. You supply your own Clerk application for auth and optionally a server-side OpenAI key. The extract does not include institute admin, billing, certification backends, or private automation — see /workflowos for the boundary summary.

Who it helps

  • Practitioners designing a deployment before writing code — map human gates, integrations, and observability early.
  • MSPs packaging discovery and design phases with exportable client artefacts.
  • CPAP candidates building portfolio evidence from real workflow exports.
  • Researchers and auditors who need a concrete PSF mapping, not slides.
Independence note: WorkflowOS is community-supported reference software — not a managed SaaS with an SLA. Production AI Institute maintains the hosted edition and publishes the extract; self-hosters own their deployment operations.

Quick start

git clone https://github.com/troygamble/workflowOS.git
cd workflowOS
npm install
cp .env.example .env.local
npm run dev

Read the full README on GitHub for Clerk setup and AI route configuration. For integrator programme context, see What Is a Certified AI Integrator? and the MSP programme.

Apply the standard

Turn the evidence into production practice.

Use the PSF, research library, and Lab material to review your own deployment. Credentials are available when a client, employer, or regulator needs public proof.

The Production AI Brief