Specflow turns vague product ideas into clear specs, technical plans, and PRs ready for human review — powered by AI agents that understand your codebase.
Roles detected
How it works
Start with raw product language. An AI analyst asks the hard questions, surfaces edge cases, and eliminates ambiguity before a single line of code.
PO: "Add multi-step onboarding with role-based flows"
Agent: What roles? Admin, Member, Viewer?
Agent: Should incomplete onboarding block access?
Agent: Email verification before or after role selection?
You get a structured spec: acceptance criteria, constraints, every open question answered. You review it and approve before anything gets built.
## Acceptance Criteria
✓ 3 roles with distinct onboarding paths
✓ Progress persistence across sessions
✓ Email verification at step 2
✓ Incomplete onboarding → limited access
An AI architect reads the spec and your actual codebase. It proposes milestones, file changes, test strategy. The developer refines and approves.
Milestone 1: DB schema + migrations
Milestone 2: Role-based flow components
Milestone 3: Progress persistence API
Milestone 4: Integration tests + e2e
Files: 12 modified · 4 new · 1 migration
An autonomous agent implements the approved plan in an isolated environment. It tests, lints, and verifies — then opens a PR ready for your review.
$ cloning repo → feat/onboarding-workflow
$ applying milestone 1/4 … done
$ running tests … 47/47 passed ✓
$ running linter … 0 warnings ✓
$ opening PR #142 … ready for review
The engine room
Each piece is independent, open source, and replaceable. Together they form a system where humans decide and machines execute.
The Interface
Where product ideas become precise specifications and technical plans. A web app that turns messy ideas into structured specs through AI-guided conversations.
The Brain
The AI agent runtime that orchestrates models, tools, and context. Fast, secure, and built to be extended — the engine that powers every automated step.
The Executor
Receives approved plans and runs them autonomously in Docker containers. Writes code, runs tests, opens PRs. Fire and forget.
Why Specflow
Small teams ship like large ones. Define specs in the morning, refine plans in the afternoon, review PRs the next day.
No more ping-pong between product and engineering. Specs are complete before implementation starts.
Every PR links back to its spec and plan. Know exactly why every line of code exists.
You decide the what and why. Agents handle the how. Architecture stays intentional.
"Remove ambiguity first.
Then automate at full speed."
We're opening early access to a small group of developer teams. Join the list — your feedback shapes what gets built next.