Free launches are open: 12 a day, and open-source projects get a dofollow link with no badge.

Submit

spec-first

Specification-driven development for AI coding agents

Developer ToolsOpen sourceOpen source · verified#ai-coding#agentic-development#version-control

Listed by the ProductPerch team from Hacker News · not launched here yet

What is spec-first?

Spec-first is a specification-driven workflow for AI coding agents. Every change begins as a written specification stored in the repository. Before any code is generated, a script verifies that every cited file and line actually exists. A separate reviewer agent then examines the specification against the codebase and either approves it, requests cuts, or refuses it.

Once approved, the main agent builds the feature on its own branch and worktree. It follows the specification steps, writes the required tests, and logs every decision that was not obvious in the spec. A git commit message hook runs the project build and tests as a gate. Only when these checks pass does the branch push and open a pull request containing the original specification.

The tool ships as a Claude Code plugin but functions with Codex, Cursor, and Copilot using plain YAML and Markdown templates. Each specification receives an isolated environment, allowing multiple agents to work in parallel without conflicts. State is preserved in context files so work resumes after interruptions. The project is MIT licensed and fully open source.

spec-first features

  • Writes specifications before generating code
  • Verifies file and line citations against existing code
  • Runs a separate reviewer agent to approve or modify specs
  • Builds features in isolated branches and worktrees
  • Gates commits with project build and test checks
  • Logs every non-obvious decision made during generation
  • Supports parallel execution across multiple agents

Who uses spec-first and for what

  • Guiding AI coding agents to implement features without drifting from requirements
  • Maintaining code quality by requiring review before implementation
  • Running multiple AI development tasks simultaneously in isolated environments
  • Enforcing CI/CD standards automatically for all AI-generated commits
  • Tracking and auditing decisions made by AI agents during development

spec-first FAQ

Is spec-first free?

Yes, the project is MIT licensed and available at no cost. The website does not mention paid tiers or subscriptions.

Does spec-first work with Claude Code?

Yes, it installs directly as a Claude Code plugin using the marketplace command. It also provides templates for other agents.

Can I use spec-first with Cursor or Copilot?

Yes. You can copy the provided YAML and Markdown templates into your project, configure AGENTS.md, and run the installation script to set up the commit gate.

Is spec-first open source?

Yes, the source code is publicly available on GitHub under the MIT license.

Does spec-first require a backend server?

No, it operates locally within your repository using plain files, Git hooks, and agent commands. The website does not mention any cloud dependencies or required servers.

Comments (0)

Comments and upvotes open when this product launches on ProductPerch.

More Developer Tools launches

All Developer Tools tools

Launch packs

Buy paid launches