The six pillars

fledge organizes the dev workflow into six pillars. They cover the whole lifecycle from project creation to release, and they’re explicitly the only things core ships. Anything else is a plugin.

Scaffold --> Run --> Spec --> AI --> Ship
                                       \
                  Extend (the protocol that lets plugins add the rest)

Scaffold

Get a project off the ground. Pick a template (built-in or remote), scaffold it, and you’re writing code in under a minute.

Commands: templates init, templates create, templates validate, templates list, templates search, templates publish

Run

Define your tasks, wire them into pipelines, watch files for re-runs. This is where fledge.toml lives.

Commands: run, lanes, watch

Spec

spec-sync. Every module declares a contract (specs/<name>/<name>.spec.md plus optional companion files). The contract is the source of truth for why a module exists, and AI commands inject the relevant specs as context. Code and docs literally cannot drift.

Commands: spec (subcommands: check, init, list, show, new)

AI

Provider-agnostic AI in the daily-driver path. Switch between Claude CLI and any Ollama-speaking endpoint in one line. Ask questions about your codebase. Review your diff with one model or a panel of them in parallel.

Commands: ai (status, models, use), ask, review

Ship

Branch, draft an AI-written PR body, preview, confirm, push. Then bump version, generate changelog, tag, push the tag.

Commands: work (start, commit, push, status), release, changelog

For GitHub-specific browsing of the resulting PR, checks, and issues, install fledge-plugin-github. Ships in the default plugin set.

Extend

Plugins, configuration, command-tree introspection, environment diagnostics, shell completions. The mechanism layer.

Commands: plugins, config, introspect, completions, doctor

fledge plugins install --defaults installs the curated set:

Why these are plugins and not core: each one bakes an ecosystem assumption (GitHub-only, polyglot lockfile parsers, niche metrics) that not every fledge user needs. Plugins keep the binary small and let each capability evolve independently.