Skip to main content

Modern CLI installer for nwave with rich terminal UI

Project description

nWave

AI agents that guide you from idea to working code — with you in control at every step.

nWave runs inside Claude Code. You describe what to build. Specialized agents handle requirements, architecture, test design, and implementation. You review and approve at each stage.

Quick Start

Choose your installation method below. Both work equally well; the plugin method has zero dependencies and is recommended for most users.

Plugin (Recommended — 1 command, zero dependencies)

From Claude Code, run:

/plugin marketplace add nwave-ai/nwave
/plugin install nwave@nwave-marketplace

That's it. Restart Claude Code and type /nw: to see all available commands.

Coming soon: We're working to be hosted on Anthropic's official plugin marketplace. Once available, installation will be even simpler:

/plugin install nwave@claude-plugins-official

CLI Installer (Python 3.10+ required)

For users who prefer pip-based installation or need advanced configuration:

pipx install nwave-ai
nwave-ai install

No repository clone needed. This installs nWave from PyPI and sets up agents and commands in ~/.claude/.

Don't have pipx? Install it first: pip install pipx && pipx ensurepath, then restart your terminal. pipx docs. Windows users: Use WSL, not cmd.exe or PowerShell. Install WSL first: wsl --install

Full setup details: Installation Guide

Which method should I use?

I am... Use Why
Trying nWave for the first time Plugin Zero dependencies, instant setup
Rolling out to my team Plugin Same command on every machine
Already using the CLI installer Either Both coexist safely, migrate at your pace
Contributing to nWave development CLI Access to dev scripts and DES internals

Use (inside Claude Code, after reopening it)

/nw:discuss "user login with email and password"   # Requirements
/nw:design --architecture=hexagonal                 # Architecture
/nw:distill "user-login"                            # Acceptance tests
/nw:deliver                                         # TDD implementation

Four commands. Four human checkpoints. One working feature.

Full walkthrough: Your First Feature

Staying Updated

nWave checks for new versions when you open Claude Code. When an update is available, you'll see a note in Claude's context with the version and what changed.

If you used the plugin method:

/plugin marketplace update nwave-marketplace

Or enable auto-updates in your Claude Code plugin settings.

If you used the CLI installer:

pipx upgrade nwave-ai   # pull the latest package (use pip install --upgrade nwave-ai if you installed via pip)
nwave-ai install        # deploy the new framework files to ~/.claude/

To control check frequency: set update_check.frequency in ~/.nwave/des-config.json (daily / weekly / every_session / never).

Control Your Token Spend — Without Sacrificing Quality

nWave enforces proven engineering practices — TDD, peer review, mutation testing — at every step. /nw:rigor lets you scale the depth of those practices to match the stakes of your work. A config tweak doesn't need the same investment as a security-critical feature. You pick the profile; nWave enforces it everywhere.

/nw:rigor                    # Interactive: compare profiles, pick one
/nw:rigor lean               # Quick switch: apply immediately
/nw:rigor custom             # Build your own: choose each setting
Profile Agent Reviewer Review TDD Mutation Est. Cost When to Use
lean haiku -- no R→G no lowest Spikes, config, docs
standard sonnet haiku single full 5-phase no moderate Most work
thorough opus sonnet double full 5-phase no higher Critical features
exhaustive opus opus double full 5-phase ≥80% kill highest Production core
custom you choose you choose you choose you choose you choose depends Your exact combo
inherit yours haiku single full 5-phase no varies You pick the model

Pick once — it persists across sessions. Change anytime. Every /nw:deliver, /nw:design, /nw:review respects your choice automatically. Need a combination no preset covers? /nw:rigor custom walks you through each setting.

/nw:rigor lean        # prototype fast
/nw:deliver           # haiku crafter, no review, RED→GREEN only
/nw:rigor standard    # ready to ship — bump up
/nw:deliver           # sonnet crafter, haiku reviewer, full TDD

Understanding DES Messages

DES is nWave's quality enforcement layer — it monitors every agent Task invocation during feature delivery to prevent unbounded execution, enforce TDD discipline, and protect accidental edits. Most DES messages are normal enforcement, not errors. They appear when agents skip required safety checks or when your code contains patterns that look like step execution.

DES also runs automatic housekeeping at every session start: it removes audit logs beyond the retention window, cleans up signal files left by crashed sessions, and rotates the skill-loading log when it grows too large. This happens silently in the background and never blocks your session.

Message What It Means What To Do
MISSING_MAX_TURNS Task invocation forgot to set max_turns parameter. Add max_turns=30 (or appropriate value) to the Task call. Recommended: 15 (quick), 25 (background), 30 (standard), 35 (research).
DES_MARKERS_MISSING Task prompt mentions a step ID (01-01 pattern) but lacks DES markers. Either: add DES markers for step execution, OR add <!-- DES-ENFORCEMENT : exempt --> comment if it's not actually step work.
Source write blocked You tried to edit a file during active /nw:deliver outside a DES task. Edit requests must go through the active deliver session. If you need to make changes, finalize the current session first.
TDD phase incomplete Sub-agent returned without finishing all required TDD phases. Re-dispatch the same agent to complete missing phases (typically COMMIT or refactoring steps).
nWave update available SessionStart detected a newer version available. Optional. Run pipx upgrade nwave-ai && nwave-ai install when ready to upgrade, or dismiss and continue working.
False positive blocks Your prompt accidentally matches step-ID pattern (e.g., dates like "2026-02-09"). Add <!-- DES-ENFORCEMENT : exempt --> comment to exempt the Task from step-ID enforcement.

These messages protect code quality but never prevent your work — they guide you toward the safe path.

How It Works

  machine        human         machine        human         machine
    │              │              │              │              │
    ▼              ▼              ▼              ▼              ▼
  Agent ──→ Documentation ──→ Review ──→ Decision ──→ Agent ──→ ...
 generates    artifacts      validates   approves    continues

Each wave produces artifacts that you review before the next wave begins. The machine never runs unsupervised end-to-end.

The full workflow has six waves. Use all six for greenfield projects, or jump straight to /nw:deliver for brownfield work.

Wave Command Agent Produces
DISCOVER /nw:discover product-discoverer Market validation
DISCUSS /nw:discuss product-owner Requirements
DESIGN /nw:design solution-architect Architecture + ADRs
DEVOPS /nw:devops platform-architect Infrastructure readiness
DISTILL /nw:distill acceptance-designer Given-When-Then tests
DELIVER /nw:deliver software-crafter Working implementation

22 agents total: 6 wave agents, 5 cross-wave specialists, 11 peer reviewers. Full list: Commands Reference

Documentation

Getting Started

Guides & Reference

Community

Privacy

nWave does not collect user data. See Privacy Policy for details.

License

MIT — see LICENSE for details.

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nwave_ai-1.5.0rc1.tar.gz (2.6 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

nwave_ai-1.5.0rc1-py3-none-any.whl (954.3 kB view details)

Uploaded Python 3

File details

Details for the file nwave_ai-1.5.0rc1.tar.gz.

File metadata

  • Download URL: nwave_ai-1.5.0rc1.tar.gz
  • Upload date:
  • Size: 2.6 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for nwave_ai-1.5.0rc1.tar.gz
Algorithm Hash digest
SHA256 4d1e4d9a1468c14906deaf7dca8575766a34bf08b99c86b836c8be58f5bdb1fb
MD5 c0ece1987ef402182e8d132d557cdfc9
BLAKE2b-256 25bd3152337d30c85435025a82dec4641de73d0aa849115c57d309e96380a05e

See more details on using hashes here.

Provenance

The following attestation bundles were made for nwave_ai-1.5.0rc1.tar.gz:

Publisher: release-rc.yml on nWave-ai/nwave-dev

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file nwave_ai-1.5.0rc1-py3-none-any.whl.

File metadata

  • Download URL: nwave_ai-1.5.0rc1-py3-none-any.whl
  • Upload date:
  • Size: 954.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for nwave_ai-1.5.0rc1-py3-none-any.whl
Algorithm Hash digest
SHA256 5019a74ba5ad7b01b331e0b3c52a3bfb7547b7cbc10351e62d6b2b1d7c4460c0
MD5 1afe8bd610a62f990dff953ce78293f7
BLAKE2b-256 b7727626308e46972ccfe41da465377f5f96357338c2284d9b25d1f7af8b7c29

See more details on using hashes here.

Provenance

The following attestation bundles were made for nwave_ai-1.5.0rc1-py3-none-any.whl:

Publisher: release-rc.yml on nWave-ai/nwave-dev

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page