The structured AI development toolkit that turns ideas into executable specifications.
Phase 0 AI ideation · Enterprise governance · 18 AI coding agents · Upstream-compatible.
- What This Project Is
- Why This Project Exists
- Core Capabilities
- Workflow at a Glance
- Supported AI Agents
- What is
uv? - Install
- Quick Start
- CLI Reference
- Extension System
- Repository Structure
- Compatibility and Upstream Sync
- Documentation Index
- Development and Testing
- Security, Support, and Governance
- Roadmap
- License
Nexo Spec Kit is a structured AI development toolkit and CLI for teams who build software with AI agents. It implements a complete Spec-Driven Development (SDD) pipeline that starts before any code is written:
- Phase 0 — Ideate: Structured AI ideation (IDEATE → SELECT → STRUCTURE → VALIDATE) to discover what to build before writing a single spec.
- Phase 1 — Specify: Author executable specifications that constrain AI agents to the right scope.
- Phase 2 — Plan and implement: Generate traceable implementation plans and delegate execution to AI agents.
- Phase 3 — Validate: Automated scope gates, contract validation, and handoff metadata create an auditable trail from idea to shipped code.
Nexo Spec Kit maintains optional upstream compatibility with
github/spec-kit. Set SPECIFY_TEMPLATE_REPO=github/spec-kit or use
--template-repo github/spec-kit to use upstream templates. This toolkit is
not affiliated with GitHub. See docs/trademarks.md.
For the full product identity and strategic positioning, see
docs/positioning.md.
AI coding agents are excellent at implementing but poor at deciding what to build. Without a structured front-end discovery phase, teams accumulate poorly-scoped features, duplicated agent context, and brittle handoffs.
Nexo Spec Kit closes this gap by adding a complete Phase 0 ideation layer on top of the Spec-Driven Development workflow, plus enterprise-grade governance that makes every AI decision traceable and auditable:
- Phase 0 AI ideation workflow (
ideate → select → structure → validate) - 18 AI coding agent integrations (the same named agents as upstream)
- 5 additional CLI commands vs upstream (
hierarchy-contract,scope-detect,scope-gate,productivity,update) - contract validation, scope gates, and handoff metadata
- enterprise compliance checker and operational release/deploy workflows
specify initproject bootstrap with:- single or multi-agent selection (
--ai claudeor--ai copilot,claude) - script type control (
--script sh|ps) - dry-run and automation flags (
--dry-run,--no-banner,--no-git) - template source override (
--template-repo,SPECIFY_TEMPLATE_REPO)
- single or multi-agent selection (
- Phase 0 slash-command set:
/speckit.brainstorm(optional strategic pre-phase that feeds IDEATE)/speckit.execution-plan(optional strategic roadmap artifact before readiness review)/speckit.strategic-review(optional strategic readiness gate before public launch)/speckit.ideate/speckit.select/speckit.structure/speckit.validate
- SDD slash-command set:
/speckit.constitution/speckit.specify/speckit.clarify/speckit.plan/speckit.tasks/speckit.implement/speckit.amend/speckit.analyze/speckit.checklist/speckit.taskstoissues
- Runtime utility commands:
specify checkspecify versionspecify update
- Extension management:
specify extension list|add|remove|search|info|update|enable|disable
IDEATE -> SELECT -> STRUCTURE -> VALIDATE
Optional strategic pre-step:
BRAINSTORM -> EXECUTION-PLAN -> IDEATE -> SELECT -> STRUCTURE -> VALIDATE
Optional pre-launch strategic quality gate:
... -> IMPLEMENT -> STRATEGIC-REVIEW
Purpose:
- explore alternatives early
- score and select a direction
- capture vision artifacts
- gate feasibility before implementation planning
CONSTITUTION -> SPECIFY -> CLARIFY -> PLAN -> TASKS -> IMPLEMENT
Purpose:
- encode project principles
- define what to build before how to build
- generate implementation artifacts and actionable tasks
specify init --ai <agent> supports the following values:
--ai value |
Agent | Type | Output Folder |
|---|---|---|---|
copilot |
GitHub Copilot | IDE | .github/ |
claude |
Claude Code | CLI | .claude/ |
gemini |
Gemini CLI | CLI | .gemini/ |
cursor-agent |
Cursor | IDE | .cursor/ |
qwen |
Qwen Code | CLI | .qwen/ |
opencode |
opencode | CLI | .opencode/ |
codex |
Codex CLI | CLI | .codex/ |
windsurf |
Windsurf | IDE | .windsurf/ |
kilocode |
Kilo Code | IDE | .kilocode/ |
auggie |
Auggie CLI | CLI | .augment/ |
codebuddy |
CodeBuddy | CLI | .codebuddy/ |
qoder |
Qoder CLI | CLI | .qoder/ |
roo |
Roo Code | IDE | .roo/ |
q |
Amazon Q Developer CLI | CLI | .amazonq/ |
amp |
Amp | CLI | .agents/ |
shai |
SHAI | CLI | .shai/ |
agy |
Antigravity | IDE | .agent/ |
bob |
IBM Bob | IDE | .bob/ |
For per-agent details and install links, see docs/agents.md and the live AGENT_CONFIG in src/specify_cli/__init__.py.
uv is an extremely fast Python package and project manager, written in Rust. It's a modern replacement for tools like pip, pip-tools, pipx, poetry, pyenv, and virtualenv — all in a single, blazingly fast tool.
Spec Kit uses uv because it:
- Installs and manages Python versions automatically
- Installs Python packages 10-100x faster than pip
- Provides a built-in tool manager (replacing
pipx) for CLI tools likespecify - Works consistently across macOS, Linux, and Windows
- Requires zero configuration to get started
curl -LsSf https://astral.sh/uv/install.sh | shpowershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"brew install uvsudo snap install astral-uv --classicpip install uvFor more installation options and troubleshooting, see the official uv installation guide.
After installation, verify that uv is available:
uv --version- Python
3.11+ uv(recommended package/runtime manager)- Git
- AI agent of your choice from the table above
ripgrep(rg) recommended for validator scriptsyqrecommended for broader script/tooling compatibility checks
uvx --from git+https://github.com/nsalvacao/spec-kit.git specify init my-project --ai codexuv tool install specify-cli --from git+https://github.com/nsalvacao/spec-kit.git
specify --helppipx install git+https://github.com/nsalvacao/spec-kit.git
# or
pip install git+https://github.com/nsalvacao/spec-kit.gitspecify init my-project --ai codex
cd my-projectUse current directory mode if needed:
specify init . --ai claude --script shWhen .specify/ already exists, init preserves it by default:
specify init . --ai geminiUse --force only when you explicitly want to reinitialize and overwrite .specify:
specify init . --ai gemini --forceInitialize with Antigravity support:
specify init my-project --ai agy/speckit.ideate
/speckit.select
/speckit.structure
/speckit.validate
/speckit.constitution
/speckit.specify
/speckit.clarify
/speckit.plan
/speckit.tasks
/speckit.implement
/speckit.amend
/speckit.checklist
/speckit.analyze
/speckit.taskstoissues
specify updateIf codex is selected, init also prints a CODEX_HOME setup hint for project-scoped prompts.
| Command | Description |
|---|---|
specify init |
Initialize project from templates and generate agent command packs |
specify check |
Validate required/recommended tools and agent CLIs |
specify version |
Show CLI, template, and environment version info |
specify update |
Check for updates and optionally upgrade CLI |
specify extension ... |
Manage extensions in .specify/extensions/ |
| Option | Purpose |
|---|---|
--ai |
Select one or more agents (a,b,c) |
--script sh|ps |
Force POSIX or PowerShell script variant |
--template-repo |
Override template source repo (owner/name) |
--here / . |
Initialize in current directory |
--force |
Skip prompts and overwrite existing .specify content when reinitializing |
--ignore-agent-tools |
Skip CLI tool checks for selected agents |
--dry-run |
Preview without writing files |
--no-banner |
Suppress ASCII banner (CI-friendly) |
--no-git |
Skip git init |
--github-token |
Use token for GitHub API operations |
--skip-tls |
Disable TLS verification (local troubleshooting only) |
--local-templates |
Path to local .genreleases dir for development testing (bypasses GitHub download) |
This fork includes an extension framework and CLI lifecycle commands.
specify extension list
specify extension search
specify extension add <name-or-path>
specify extension info <name>
specify extension remove <name>Notes:
- extension commands must run from a spec-kit project root (
.specify/present) - local/dev and URL-based installation are supported
- catalog and extension templates live under
extensions/
See:
extensions/EXTENSION-USER-GUIDE.mdextensions/EXTENSION-DEVELOPMENT-GUIDE.mdextensions/RFC-EXTENSION-SYSTEM.md
spec-kit/
├── src/specify_cli/ # CLI implementation
├── templates/ # Slash-command and artifact templates
├── scripts/ # Bash, PowerShell, and Python automation scripts
├── docs/ # Markdown docs source
├── docs-site/ # MkDocs site and published docs assets
├── extensions/ # Extension framework, docs, and templates
├── tests/ # Pytest suite
├── .github/workflows/ # CI/CD pipelines
├── spec-driven.md # Methodology reference
└── AGENTS.md # Agent operating guidance for this repo
- Default template source is this fork:
nsalvacao/spec-kit - You can opt into upstream templates:
export SPECIFY_TEMPLATE_REPO=github/spec-kit
specify init my-project --ai codexOr per command:
specify init my-project --template-repo github/spec-kit --ai codexDetailed policy:
docs/compatibility.mddocs/upstream-sync.md
| Topic | Link |
|---|---|
| Product positioning | docs/positioning.md |
| ADR — Positioning model | docs/adr-positioning-model.md |
| Getting started | docs/quickstart.md |
| Installation | docs/installation.md |
| Configuration | docs/configuration.md |
| Scope scoring rubric | docs/scope-scoring-rubric.md |
| Scope gate contract | docs/scope-gate-consumption-contract.md |
| CLI reference | docs/cli.md |
| Agent support | docs/agents.md |
| Methodology | docs/methodology.md |
| Walkthrough | docs/walkthrough.md |
| Upgrade guidance | docs/upgrade.md |
| Release process | docs/release.md |
| Distribution | docs/distribution.md |
| Compatibility | docs/compatibility.md |
| Upstream sync | docs/upstream-sync.md |
| GitHub Pages site | https://nsalvacao.github.io/spec-kit/ |
uv sync
uv run specify --helpuv run pytest tests/ --tb=short -vpip install mkdocs-material
mkdocs build -f docs-site/mkdocs.yml -d docs-site/siteMore details:
CONTRIBUTING.mddocs/local-development.md
- Security policy:
SECURITY.md - Support process:
SUPPORT.md - Code of conduct:
CODE_OF_CONDUCT.md - Contribution rules and phase taxonomy:
CONTRIBUTING.md
If you use AI assistance in contributions, disclosure is required (see CONTRIBUTING.md).
This fork is an independent derivative of github/spec-kit distributed under the MIT license.
Trademark usage is attribution-only; this fork is not affiliated with GitHub.
- License and copyright obligations:
docs/legal-compliance.md - Upstream intake provenance workflow:
docs/provenance-playbook.md - Trademark-safe usage guidance:
docs/trademarks.md - Upstream synchronization policy:
docs/upstream-sync.md
Compliance is enforced automatically via the compliance-guard CI workflow on every
push and pull request.
Roadmap and evolution priorities:
ROADMAP.md- open issues:
https://github.com/nsalvacao/spec-kit/issues
MIT. See LICENSE.