REVIEWED & READY TO INSTALL
Spec Kit
github/spec-kit
Spec Kit is GitHub's open source toolkit that makes AI coding assistants (Claude Code, GitHub Copilot, Cursor, Gemini CLI, Codex and 40+ more) follow a process: write the spec, plan, split into tasks, build, then check the code against the spec. Optional add-ons for bug fixing and idea assessment. Free, MIT license.
STEP BY STEP
How to install Spec Kit
The step-by-step guide is right on this page, written for people who have never installed anything from GitHub.
Updated 11/21/2026SIGNUP BONUS
Sign up and get 5 xu to try the AI tools
Every new GitHot account gets 5 xu — enough to run the AI Repo Explainer, MCP Config Generator or Repo Health Check. Free, just an email or Google.
- GitHub stars
- 150,949
- Language
- Python
- Trending rank
- #7
- Added
- 11/21/2026
How to install Spec Kit, step by step
Work from top to bottom. Each grey box is one command: copy the whole line, paste it into the terminal window and press Enter.
Never used a terminal? Read this first (1 minute)
A terminal is a window where you type commands for your computer. You do not need to understand them, just copy and paste them exactly.
- Windows: press the Windows key, type "PowerShell", then press Enter.
- macOS: press Command + Space, type "Terminal", then press Enter.
- Linux: press Ctrl + Alt + T.
- Copy a command from the guide, right-click inside the window you just opened to paste it (macOS: Command + V), then press Enter.
- Wait for the command to finish (the cursor blinks again on a new line) before running the next one.
Lines starting with # are notes; you do not need to copy them. If red error text appears, check the common errors section at the end of the guide.
1. What is Spec Kit?
Spec Kit is an open source toolkit from GitHub itself. It makes AI coding assistants (Claude Code, GitHub Copilot, Cursor, Gemini CLI, Codex and many more) follow a process instead of guessing what you meant and writing code straight away. Its main process is called Spec-Driven Development: you first describe what you want and why, the assistant turns that into a specification (spec), and only then does it plan the technical approach, split the work into tasks, write the code and check the code against the spec.
Spec Kit has two parts: a command-line program called Specify, which creates projects and installs the commands into your assistant, and a set of /speckit-... commands that you type in the assistant's chat. Besides building new features, it ships two optional add-ons: verified bug fixing, and assessing whether an idea is worth building at all.
It is free and MIT licensed. Version 1.0 came out on 21 August 2026; the latest release when this guide was written is v1.4.3, published on 19 November 2026.
To be clear: this is a tool for people who already use AI to build software. Spec Kit has no app window of its own and does not write code by itself. It runs inside an AI coding assistant you already have.
A good fit if:
- You use an AI coding assistant and it often misreads the request, leaves features out or wanders off.
- You want to build a small or medium app with real paperwork behind it: a spec, a plan and a task list you can come back to later.
- You work in a team where people use different AI assistants but want one shared process.
Not a good fit yet if:
- You have never used an AI coding assistant. Install one, such as Claude Code or GitHub Copilot, first.
- You only need a one-line fix or a quick question. A multi-step process is overkill there.
- You do not want to read what the assistant writes. Spec Kit pays off because you review each step.
2. What you need
- A Windows, macOS or Linux computer. On Windows it runs directly in PowerShell; WSL is not required.
- Python 3.11 or newer (the programming language Specify is written in). Download it from https://www.python.org/downloads/ and check with
python --version. - uv (Astral's tool for installing Python programs; it is fast and keeps them separate from the Python already on your computer). Step 1 shows how to install it.
- An AI coding assistant that is installed and signed in. Spec Kit has ready-made integrations for more than 40 tools, including Claude Code, GitHub Copilot, Cursor, Gemini CLI, Codex CLI, Antigravity, opencode, Qwen Code, Kiro CLI, Trae and Zed. The account and its costs belong to that tool.
- Git (software that records the history of your code, https://git-scm.com/downloads) is optional. You only need it if you turn on the git extension, which gives each feature its own branch.
Note: Spec Kit needs no API key of its own. Writing specs and code is done by your AI assistant, so it uses that assistant's usage limits or tokens.
3. Step-by-step installation
Step 1: Install uv
Windows (type in PowerShell):
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Or, if your computer has WinGet:
winget install --id=astral-sh.uv -e
macOS / Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
On a Mac with Homebrew you can also use brew install uv.
When it finishes, close the terminal window and open a new one, then type:
uv --version
What you should see: a line starting with uv and a version number, for example uv 0.12.10.
Step 2: Install the Specify CLI
uv tool install specify-cli
This downloads the official specify-cli package from PyPI (the Python package index) and installs it as a specify command you can use from any folder. Check it:
specify version
What you should see: Specify prints the version you installed (the latest when this guide was written is 1.4.3).
The official installation page recommends installing from the source on GitHub, pinned to one release (here, the latest when this guide was written). Both routes are supported:
uv tool install specify-cli --from git+https://github.com/github/spec-kit.git@v1.4.3
Note: If you use pipx instead of uv, the equivalent is
pipx install specify-cli. Pick one method only; installing with two tools leaves two copies ofspecifyfighting each other.
Step 3: Create a new project and choose your AI assistant
Go to the folder where you keep your projects and run the command for your assistant. For example, with Claude Code:
specify init my-project --integration claude
Change the value after --integration to match your tool:
| AI assistant | --integration value |
|---|---|
| Claude Code | claude |
| GitHub Copilot | copilot |
| Cursor | cursor-agent |
| Gemini CLI | gemini |
| Codex CLI | codex |
| Antigravity | agy |
| opencode | opencode |
| Qwen Code | qwen |
The full list is on the official site: https://github.github.io/spec-kit/integrations/index.html . If you leave out --integration, Specify shows a list you can pick from with the arrow keys.
Specify also asks "Choose script type (or press Enter)": just press Enter to accept the default (ps, PowerShell, on Windows; sh, Bash, on macOS and Linux).
What you should see: a "Specify Project Setup" box with your project name, the lines "Selected coding agent integration" and "Selected script type", then "Project ready." in green, a yellow "Agent Folder Security" box suggesting you add the assistant's folder to .gitignore (so you never publish login details by accident), and a "Next Steps" box listing the commands constitution, specify, plan, tasks, implement and converge.
Move into the new folder:
cd my-project
Step 4 (if you already have code): add Spec Kit to an existing project
Open a terminal inside your project folder and run:
specify init --here --integration claude
If the folder already contains files, Specify asks "Template files will be merged with existing content and may overwrite existing files. Do you want to continue?". Back up or commit your code first, then type y and press Enter. To skip the question, add --force:
specify init --here --force --integration claude
What you should see: the same as Step 3, ending with "Project ready." Your project now has a .specify folder (project principles, templates and scripts) and your assistant's command folder, for example .claude/skills for Claude Code.
4. First use
Open your AI assistant inside the project folder (for example, type claude in that folder). Type the commands below into the assistant's chat, not into the terminal. Run one at a time and read the result before moving on.
A trial run, building a photo organizer:
- Set the project principles (only once):
/speckit-constitution Prefer readable code, include tests, keep the interface simple.
- Describe what to build, without naming any technology yet:
/speckit-specify Build an app that organizes photos into albums grouped by date, with a tile grid preview of each album.
- Make the technical plan; this is where you pick the stack:
/speckit-plan Use Vite with plain JavaScript. Keep images on the computer and store album data in SQLite.
- Split the work, then build it:
/speckit-tasks
/speckit-implement
- Check the code against the spec:
/speckit-converge
What you should see: after step 2, the specs/ folder holds a new folder for the feature with a spec.md file. Step 3 adds plan.md, step 4 adds tasks.md (tasks in dependency order) and the assistant starts writing code. If converge finds gaps, it appends new tasks to tasks.md; run /speckit-implement and /speckit-converge again until it reports Converged.
For important work, add three quality gates: /speckit-clarify (the assistant asks up to five questions about unclear points; run it before plan), /speckit-checklist (a checklist for the quality of the requirements) and /speckit-analyze (looks for contradictions between spec, plan and tasks before implement; it only reads, it never edits files).
Note: The command spelling depends on the assistant. Claude Code, Copilot, Cursor and Antigravity use
/speckit-specify; Codex uses$speckit-specify; Gemini CLI uses/speckit.specify. The "Next Steps" box from Step 3 always shows the right spelling for the assistant you chose.
5. Common errors and fixes
| Symptom | Cause | Fix |
|---|---|---|
uv or specify is "not recognized" / "command not found" | Not installed yet, or the terminal was opened before installing and does not see the new path | Close every terminal window, open a new one and try again. If it still fails, repeat Step 1 or Step 2 |
| Red "Agent Detection Error" box saying "... not found" | Specify cannot find the program of the assistant you chose | Install that assistant from the link Specify prints. If the assistant runs on another machine or inside an IDE, add --ignore-agent-tools to the init command |
| Red "Directory Conflict" box | A folder with that name already exists | Choose another project name, or back up and add --force to merge into the existing folder |
| "No such option: --integration" | Your Specify version is too old | Reinstall the current version: uv tool install --force specify-cli |
The assistant does not recognize /speckit-specify | The assistant was started in another folder, or before you ran init | Quit the assistant and start it again inside the project folder. Check your assistant's command spelling in section 4 |
specify version still shows the old number after reinstalling | Two copies installed by two different tools (uv and pipx) | Uninstall one (section 6) and keep a single install method |
If you are still stuck, run specify check to see which tools Specify finds on your computer, then search or report the problem at https://github.com/github/spec-kit/issues .
6. Uninstall / update
Check whether a newer version exists (read-only, changes nothing):
specify self check
Update Specify:
specify self upgrade
On Windows this command cannot replace itself; it prints a uv tool install ... --force command for you to run after it exits. The manual method for any computer:
uv tool install --force specify-cli
Update the commands inside a project (run in the project folder, change claude to your assistant):
specify integration upgrade claude
The specs you wrote in specs/ are not touched by updates.
Remove Specify from your computer:
uv tool uninstall specify-cli
Removing the program does not delete files it created in your projects. To take Spec Kit out of a project, delete the .specify folder and the speckit command folders inside your assistant's folder (for example inside .claude/skills).
7. Frequently asked questions
Does it cost anything? Spec Kit is free (MIT). You only pay for the AI assistant you use. The multi-step process uses more of your assistant's quota than asking it to write code directly; in exchange the result is checked.
Can I use it if I cannot code? You can write the "what to build" part in plain language. The plan, the tasks and the resulting code still need someone who understands a little of it to review them. Spec Kit suits people who are already comfortable with an AI coding assistant.
Do I reinstall it for every project? The specify command is installed once per computer. Each project runs specify init once to add the commands.
Can one project use several AI assistants? Yes. In an initialized project, add another one, for example Gemini CLI, with specify integration install gemini, then make it the default with specify integration use gemini. Some assistants (such as Antigravity) are not declared safe to install alongside others; check that assistant's page first.
What else can it do besides new features? There is a bug-fixing add-on (specify extension add bug, then /speckit-bug-assess, /speckit-bug-fix, /speckit-bug-test) and an idea assessment add-on (specify extension add assess) that ends with a go, needs-clarification or stop decision.
Written by GitHot and checked against the project's official documentation. If a step is wrong or unclear, message GitHot through the channels at the bottom of the page.
SEE WHY IT'S HOT
Watch the original review
Go back to the video that brought you here.
Don't miss the next repo
Every new repo gets a video review, with its install guide right here.
HISTORY
Seen on GitHub Trending
- Nov 4#7147,099