Install Spec Kit: specify-cli Setup and the /speckit-* Command Changes

Spec Kit is GitHub’s open-source toolkit for Spec-Driven Development: write the specification first, then a plan and tasks, and finally let your AI coding assistant implement the tasks. Its command line is called specify and comes from the specify-cli package.

Many tutorials online still use the old install command, the --ai flag and command names like /speckit.specify, and copying them often fails. This page follows the official documentation to show the current correct way, and lists what has changed.

When this was checked

Written against the official Spec Kit documentation, checked on 2026-10-26. The project moves quickly, so trust specify --help on your machine and the official docs over this page when they differ.

Who this is for

• People who want to try a spec-driven workflow inside an AI coding assistant such as Claude Code, GitHub Copilot, Gemini CLI, CodeBuddy CLI, Pi or Oh My Pi.
• People who installed Spec Kit from an older tutorial and now get errors or missing commands.

Prerequisites

Item	Requirement
OS	Linux, macOS, or Windows (PowerShell scripts work without WSL)
Python	3.11 or higher
Package manager	uv (recommended) or pipx
AI coding assistant	Any supported assistant; the official docs list dozens of integrations
Git	Optional, only needed when the git extension is enabled

Spec Kit itself does not need Node.js. If your assistant is installed through npm, see AI Agent CLI Installation first.

Install specify-cli

Option 1: From PyPI (shortest)

# Recommended: uv
uv tool install specify-cli

# Or pipx
pipx install specify-cli

# Or pip
pip install specify-cli

To pin a version: uv tool install specify-cli==0.12.11 (use the version from the releases page).

Option 2: From a GitHub release tag

The official docs list this as the recommended reproducible route. It needs uv:

uv tool install specify-cli --from git+https://github.com/github/spec-kit.git@vX.Y.Z

Replace vX.Y.Z with a tag from Releases. Keep the leading v, for example v0.12.11.

Verify the install

specify version
specify self check

specify version confirms the command works and prints its version. specify self check is read-only and tells you whether a newer release exists.

Initialize a project

specify init my-project --integration claude
cd my-project

--integration picks your AI assistant. Common values are claude, copilot, gemini, codebuddy, pi and omp. Useful options:

• --script sh|ps|py: choose the script type. The default is ps on Windows and sh elsewhere.
• --here --force: initialize inside an existing non-empty directory. --force acknowledges the merge warning, so commit or back up local changes first.
• --non-interactive: for CI. Without --integration it defaults to GitHub Copilot, so do not be caught out by that default.

Command names: /speckit-*

After initialization, run these in your AI assistant’s chat, not in the terminal:

Step	Command	What it does
1	/speckit-constitution	Writes the project principles that later steps are checked against; once per project
2	/speckit-specify	Turns a plain-language description of the feature into spec.md
3	/speckit-clarify	Asks about gaps in the spec and folds the answers back in
4	/speckit-plan	Generates plan.md and other design artifacts for your tech stack
5	/speckit-checklist	Generates a requirements-quality checklist
6	/speckit-tasks	Generates a dependency-ordered tasks.md
7	/speckit-analyze	Read-only check for conflicts and gaps across spec.md, plan.md and tasks.md
8	/speckit-implement	Executes tasks.md in order
9	/speckit-converge	Checks the code against the spec and appends tasks for any gaps; repeat implement and converge until it reports Converged

The shortest path is specify → plan → tasks → implement → converge. Only the full path adds constitution, clarify, checklist and analyze.

Your assistant may spell them differently

The official quickstart uses the hyphenated /speckit-* names (GitHub Copilot’s default skills mode). Older tutorials and the upgrade notes still show the dotted form such as /speckit.specify. The invocation format differs between assistants and modes, so go by the names that actually appear in your assistant’s command list after initialization.

What changed compared with older tutorials

Common in older tutorials	Now
Only uv tool install ... --from git+https://...	specify-cli is also on PyPI, installable with uv, pipx or pip; installing from a GitHub tag is still supported
specify init --ai claude	The official docs use --integration claude throughout
/speckit.specify, /speckit.plan	The quickstart uses /speckit-specify and /speckit-plan, and adds /speckit-converge
/constitution propagates changes into templates and command files	It now only updates .specify/memory/constitution.md; to restore the old propagation, run specify preset add constitution-sync
Git features always on	Now an optional extension: specify extension add git
The active feature is inferred from the Git branch	.specify/feature.json records the active feature directory

Upgrade

Upgrade the command line itself:

specify self check                 # read-only: is there a newer release?
specify self upgrade --dry-run     # preview without upgrading
specify self upgrade               # upgrade to the latest stable release
specify self upgrade --tag vX.Y.Z  # pin a release tag

Upgrade the files inside an existing project, from the project directory:

specify integration status         # list installed integrations and modified files
specify integration upgrade <key>  # run once per installed integration
specify extension update           # update all extensions

Pass --force to specify integration upgrade only when it reports locally modified managed files and you have reviewed the difference.

Common problems

• command not found: uv: install uv first, then run the install command again.
• No such option: --integration: an old specify (for example 0.1.6) is earlier on your PATH. Reinstall with the same tool you used originally, for example uv tool install --force specify-cli, then confirm with specify version.
• Two copies installed: installing with uv, pipx and pip each leaves a separate copy. Upgrade with the same tool and do not mix them.
• A PyPI install does not upgrade itself: the version stays where you installed it, so use the upgrade commands above.
• Init refuses an existing directory: once you have committed or backed up, add --here --force.

Next steps

• To decide between Spec Kit and OpenSpec, read Spec Kit vs OpenSpec.
• To see the spec workflow HagiCode itself uses, read the Introduction to OpenSpec.

Official Spec Kit documentation