@looply-cli/looply
v1.0.1
Published
<p align="center"> <img src="./assets/looply-banner.svg" alt="Looply" width="760" /> </p>
Readme
📌 Current status: see PROJECT_STATUS.md for the up-to-date product snapshot, in-progress work, and next steps.
Why Looply
AI coding agents are powerful but inconsistent. Without structure, every session starts from scratch — different tone, different conventions, different quality bar.
Looply solves this by shipping a curated, versioned set of packs — Markdown artifacts that encode your team's workflows, standards, agents, and operational context. Agents read them before producing output, so every session is calibrated the same way.
How It Works
- Install a pack into your project:
looply install - Publish the pack to your AI hosts (Codex, Claude Code)
- Work through structured workflows (
idea-to-prd→prd-to-stories→story-to-production) - Intervene when needed (
replay,run-task,run-agent,reconcile) - Repeat — packs are versioned, shared, and improved over time
Quick Start
npm install -g @looply-cli/looply
cd your-project
looply installOr try without installing:
npx @looply-cli/looply --helpFeatures
| Area | Description |
|---|---|
| Packs | engineering-base, product-base, software-delivery-suite — modular, composable via includes |
| Multi-host | Publishes the same artifact set to Codex, Claude Code, and the local desktop companion |
| Workflows | idea-to-prd, prd-to-stories, story-to-production, cloud-workload-design, platform-foundation-evolution — handoff between agents |
| Interventions | replay, run-task, run-agent, reconcile — deviate from a workflow without losing state |
| Project rules | Six categories (coding-standards, testing-requirements, security-policies, etc.) — standard defaults or custom |
| ICL guidance | In-context example layer that calibrates agent output style and quality |
| Autocomplete | Bash, Zsh, PowerShell — generated from the real CLI tree |
| Desktop companion | Local Electron app for browsing project snapshots, features, and sessions |
| Docs | Built-in VitePress portal — looply docs open or automated GitHub Pages deploy |
Usage
# Interactive install with guided questionnaire
looply install
# Explicit parameters (non-interactive)
looply install \
--host codex,claude \
--scope project \
--pack all \
--locale pt-BR \
--project-mode existing-project \
--interaction-mode balanced
# Check project status
looply status
looply status --json
# Start a workflow
looply autonomy minha-feature
# Intervene during a workflow
looply replay minha-feature --from tech-spec
looply run-task minha-feature review-code
looply reconcile minha-feature
# Validate and diagnose
looply validate
looply doctor --host codex,claude --scope project
# Manage ICL guidance
looply icl set reduced
# Open documentation
looply docs openCommands
| Command | Description |
|---|---|
| install | Install packs and publish to hosts |
| uninstall | Remove looply from the project |
| reinstall | Reinstall while preserving customizations |
| validate | Validate artifacts and contracts |
| doctor | Installation diagnostics per host/scope |
| status | Consolidated project snapshot |
| sync | Sync published state with source |
| refresh-context | Update project context and code-context |
| check-updates | Check for newer pack versions |
| upgrade | Upgrade packs |
| autonomy | Derive the next host-driven cycle for a feature |
| replay | Resume a workflow from a previous artifact |
| run-task | Record a manual task execution |
| run-agent | Record a manual agent intervention |
| reconcile | Reconcile a feature after interventions |
| list | List available workflows, agents, or tasks |
| inspect | Show details of a workflow or agent |
| integrations | Expose known integration contexts |
| icl | Manage ICL example guidance |
| docs | Generate, build, and serve documentation |
| completion | Generate and install shell autocomplete |
Workflows
| Workflow | Purpose |
|---|---|
| idea-to-prd | Transform an idea into a PRD with context and constraints |
| prd-to-stories | Break a PRD into implementable stories |
| story-to-production | Take a story from tech design through review to release |
| workflow-status | Inspect state and recommend the next step |
| cloud-workload-design | Cloud topology, async-first, queueing, and governance decisions |
| platform-foundation-evolution | Shared foundation, guardrails, pipelines, identity, and observability |
Approach
Artifact-first. Every workflow stage produces a versioned Markdown artifact. Nothing lives only in chat history.
Task-first. Workflows are composed of discrete tasks, each owned by a named agent with explicit input/output contracts.
Host-agnostic core. The artifact engine doesn't know about any specific AI host. Publishing is a separate, host-aware layer.
Codebase-first. In existing projects, the real repository is always the primary source of truth. Looply context files are accelerators — not replacements.
Incremental sync. Updates are applied by file ownership. User customizations in .looply/custom/ are never overwritten.
Progressive disclosure. Packs include checklists, templates, and decision trees that grow with the team — without requiring everyone to know everything upfront.
Documentation
Full documentation at riguelbf.github.io/looply.
Contributing
Looply is an opinionated tool. Contributions are welcome in the form of issues and pull requests. When proposing changes, start with the rationale — what problem it solves and why the current design falls short.
