Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

Fission-AI/OpenSpec

Developer Tools

A lightweight spec layer that keeps AI coding assistants and developers aligned before code is written.

73,788
5,053
TypeScript
MIT
1 year old

What it does

OpenSpec adds a spec-driven development layer to AI coding assistants. Instead of leaving requirements in chat history, each change gets its own folder containing a proposal, specs written as plain Markdown requirements and scenarios, a design document and a task checklist. Humans review the plan before implementation starts, and the AI keeps working against the same artifacts.

The workflow is built around slash commands in the assistant: /opsx:explore for thinking through an approach, /opsx:propose <name> to create the change folder, /opsx:apply to implement tasks and /opsx:archive to move the finished change into an archive and update specs. The command spelling varies by tool (/opsx-propose in Cursor and GitHub Copilot, @opsx-propose in Amazon Q, $openspec-propose in Codex), and openspec init prints the correct form.

It is aimed at individuals and teams using AI coding assistants on existing codebases. The README also describes a beta Stores feature for keeping specs in a separate repository shared across multiple code repos.

Quick start

Requirements: a supported AI coding assistant, and Node.js 22.12.0 or higher if you install through npm.

  1. Install the CLI. On macOS or Linux:

    curl -fsSL https://openspec.dev/cli/install.sh | sh
    

    On Windows PowerShell:

    irm https://openspec.dev/cli/install.ps1 | iex
    

    Alternatively use npm install -g @fission-ai/openspec@latest or brew install openspec.

  2. Initialize inside your project:

    cd your-project
    openspec init
    

    The init step sets up the openspec/ directory and generates the instructions and commands for the assistants you select.

  3. In your assistant, run /opsx:propose add-dark-mode (or the equivalent command for your tool) to generate the proposal, specs, design and tasks, then /opsx:apply to implement them. When done, /opsx:archive moves the change into openspec/changes/archive/ and updates the specs.

Full walkthroughs are in the Quickstart and installation options.

Constraints

  • The CLI supports Node.js 22.12.0+ when installed via npm; other package managers and Nix are also listed.
  • The default profile includes /opsx:explore and /opsx:propose; the expanded workflow (/opsx:new, /opsx:continue, /opsx:ff, /opsx:verify, /opsx:bulk-archive, /opsx:onboard) has to be selected with openspec config profile and applied with openspec update.
  • Command syntax differs between tools. Check Supported Tools before relying on a specific slash command.
  • OpenSpec collects anonymous telemetry: command names and version only, no arguments, paths or content. It can be disabled with openspec config set telemetry.enabled false, OPENSPEC_TELEMETRY=0 or DO_NOT_TRACK=1.
  • The README recommends high-reasoning models and clearing the context window before implementation.
  • Stores is in beta; the rest of the workflow is usable without it.
  • Licensed under MIT.

HotRepo's take

OpenSpec has climbed steadily in the daily all-language snapshots: 56,979 stars on 2026-06-27, 69,167 on 2026-09-18, and 73,781 on 2026-11-03, when it ranked 9th on the TypeScript daily list. Discussions are thin — the two Hacker News threads collected here have 2 points and 1 point respectively, with a single reply describing the 1.0 workflow as action-based and aware of which artifacts already exist. The scale of the star growth relative to that discussion volume suggests broad adoption interest rather than community evaluation in these threads.

Compared with heavier spec systems such as GitHub's Spec Kit, the project's own positioning is iteration without rigid phase gates, and its markdown specs avoid a special syntax. For readers already using AI assistants on multi-file codebases, that is a low-ceremony way to make planning reviewable; the trade-off is that quality still depends on the model and on how carefully the human reviews each proposal.

Written by AI from the project's README, updated . The project's own docs are the source of truth.

Hacker News

OpenSpec stable v1.0 – with dynamic workflows
▲ 1 · 1 comment · January 26, 2026