Epic FlowState

An open-source project management platform and development standard for teams working with AI agents.

The problem

AI coding agents are fast. They write code that works. But there's no structure. No specs. No continuity between sessions. You're letting a smart intern ship to production with no review.

We spent decades building software engineering practices for a reason. Specs, code review, TDD --- they exist because shipping fast without technique is just flailing. Speed without precision creates technical debt at an alarming rate.

Most developers using AI agents are chasing output. Prompt, generate, ship, repeat. No position. No control. The tools are powerful, but the process around them hasn't caught up.

The rationale

The AI didn't need to get smarter. We needed to get better at coaching it.

Once you enforce fundamentals --- write tests first, plan before you build, review before you ship --- the quality of what AI agents produce goes through the roof. The same principles that make a 50-person engineering team produce reliable software apply when your team includes AI agents.

FlowState was built on a simple insight: discipline scales, sloppiness doesn't.

Instead of producing 10,000 lines of code a day, the goal is to produce the highest quality code following industry best practices. Every single time. Not more speed --- more precision.

The FlowState Standard

Every feature flows through four phases with approval gates between each:

Requirements (WHAT) --- Define user stories, acceptance criteria, and constraints. No code until stakeholders agree on what we're building.

Design + UX (HOW) --- Technical architecture, data models, and wireframes. The blueprint before the build.

Tasks (STEPS) --- Break implementation into discrete, estimable work. Dependencies mapped. Test requirements defined.

Implementation (BUILD) --- TDD. Write failing tests first. Implement to pass. Refactor for quality. Document everything.

The standard covers test-driven development, functional programming, TypeScript strict mode, WCAG accessibility, security by design, 80%+ test coverage, and 100% TSDoc coverage on public APIs.

Why FlowState exists

AI agents don't push back on vague requirements. They don't ask clarifying questions unless the workflow forces them to. They don't care about quality unless someone defines what quality means.

You can give an AI agent the best skills, the strictest standards, the most detailed specs. But if it doesn't understand the whole picture --- the product vision, the architecture decisions, the security requirements --- you're giving a sharp sword to someone fighting blindfolded.

The more context you give, the sharper the sword becomes.

FlowState manages the entire development lifecycle. Projects, milestones, tasks, specs, approvals --- all tracked in one system. Steering documents define your product vision, architecture, quality standards, and security requirements. AI agents read these documents and operate within those boundaries.

Built for everyone

FlowState brings in all stakeholders. Product managers write requirements. Designers review wireframes. Stakeholders approve specs through quality gates. Nobody needs to write code to participate in the process.

The best engineering teams aren't great because of their developers. They're great because everyone --- product, design, QA, leadership --- is aligned on what "done" means before a single line of code is written.

The FlowState Standard is that alignment, codified. Not AI replacing developers. Not developers winging it with AI. A disciplined team --- humans and AI --- following a standard that produces work worth shipping.