# Spec-First AI Delivery Stack

> A disciplined stack for turning approved product requirements into small, tested, AI-assisted releases with accountable human review.

- Canonical: https://gptnavi.com/stacks/spec-first-ai-delivery-stack
- Estimated monthly cost: $20-$250/engineer/month
- Last materially updated: 2026-08-19
- Best for: Product engineers, Technical founders, Startup teams

## Problems this stack solves

- Unclear requirements
- Agent scope creep
- Thin test evidence
- Slow delivery
- Unmeasured releases

## Recommended tools

### Work definition

Recommended: Linear, Notion

Why: Keep the user outcome, acceptance criteria, risks, and non-goals visible before implementation.

### Spec-driven implementation

Recommended: Kiro, Claude Code

Why: Convert approved scope into a plan and contained implementation work with repository context.

### Developer execution

Recommended: Warp, Trae

Why: Run and share repeatable implementation, test, and debugging tasks.

### Release evidence

Recommended: Vercel, PostHog

Why: Use a normal deployment path and monitor the customer behavior that justified the release.

## Beginner setup plan

1. Start with a small bug fix or one contained feature.
2. Require accepted criteria before an AI agent edits code.
3. Keep the test evidence attached to the delivery ticket.
4. Assign a person, not an agent, as release owner.

## Included workflows

- https://gptnavi.com/workflows/spec-first-ai-feature-delivery
- https://gptnavi.com/workflows/agentic-engineering-delivery-loop
- https://gptnavi.com/workflows/product-idea-to-coding-feature-spec
- https://gptnavi.com/workflows/llm-evaluation-before-production