Skip to content
Get packs

Spec Kit

A toolkit that gives AI coding agents structured, repeatable processes for building, fixing bugs, and assessin

GitHub Download v1.2.5
Stars
145k
Forks
13k
Licence
MIT
Language
Python
Last update
6 days ago
Type
Dev tool

What it is

Spec Kit is an open source toolkit that adds structure to AI coding agents. It provides ready-made processes, reusable templates, and documented outcomes so an agent works from a written specification rather than a loose prompt. Its main process, Spec-Driven Development, moves from a spec to a plan, tasks, and code, and bundled extensions cover bug fixing and idea assessment.

  • Create a project constitution, then specify, plan, and implement a new feature through agent chat commands
  • Run a structured bug workflow that assesses a cause, applies a fix, and records a verdict of verified, partial, or failed
  • Assess whether an idea deserves investment using gathered evidence before writing code
  • Adapt templates or terminology to your team's methodology using presets
  • Install a curated setup for a role or team using bundles

Great for

  • Developers who use an AI coding agent and want more predictable results from it
  • Teams that want written specifications and plans to guide agent-written code
  • Developers who want to adapt or build their own agent workflows using its templates and extensions
  • Anyone fixing a bug who wants the agent to assess the cause and verify the fix

Not ideal for

  • People who do not use a supported AI coding agent, since the toolkit is built around one
  • Developers who want a quick one-off script and do not want to add a process layer to their project
  • Beginners with no coding background, because setup happens in a terminal and the process assumes some software knowledge
Step-by-step guide: how to use Spec KitInstall, first run, examples, tips and FAQs
Skills vs Spec KitOpenSpec vs Spec Kit

How to get it

  1. 1

    Install with a package manager

    You'll need: Python 3.11+ · uv · A supported AI coding agent

    Install the Specify CLI

    uv tool install specify-cli

    Create a project using GitHub Copilot

    specify init my-project --integration copilot

    Enter the project directory

    cd my-project

    Install the bug-fixing extension (optional)

    specify extension add bug
  2. 2

    Run it

    • Launch your coding agent in the project directory
    • Run /speckit-constitution to set project principles
    • Run /speckit-specify, /speckit-plan, /speckit-tasks, and /speckit-implement in order, reviewing each result
    • Repeat implement and converge until convergence reports Converged
  3. 3

    Get the most out of it

    • Use the SDD walkthrough to learn clarification, checklists, and consistency analysis
    • Add presets to adapt templates or enforce organizational standards
    • Use workflows to automate prompts, gates, and shell tasks
    • Use project-local overrides to change a template for one project only
    • Use the artifact inventory to see which layer supplies a command or template

Downloads

Source code (.zip)main branch · straight from GitHubRelease v1.2.5Release notes and sourceOfficial website & docsgithub.github.com/spec-kit/

Strengths

  • Gives AI agents a repeatable structure instead of ad hoc prompting
  • Works with multiple coding agents through integrations, not just one
  • Extensible through extensions, presets, workflows, and catalogs

Trade-offs

  • Depends on a supported AI coding agent, so it is not useful on its own
  • Community catalog components are not audited, so you must review source code before installing
  • The README is truncated, so some setup details and requirements are not fully covered here

How it compares

ProjectStarsTypeLevelUpdatedBest for
spec-kit (this)145kDev toolintermediate6 days agoDevelopers who use an AI coding agent and want more predictable results from it
OpenSpec73.7kDev toolbeginner4 days agoDevelopers who use AI coding assistants and want to agree on requirements before code is written
9router31.8kDev toolintermediate9 days agoDevelopers who use AI coding tools and want to reduce spending
OmniRoute83.8kAppintermediate4 days agoDevelopers who use several AI coding tools and want one place to configure providers
standard29.4kDev toolbeginner1.3 years agoJavaScript projects where the team wants consistent style without debating rules
compound-engineering-plugin25.9kAppbeginner4 days agoDevelopers already using Claude Code, Cursor, or Codex who want structured planning and review steps

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.