m6t (m‑anifes‑t) is a cross-platform desktop workbench for managing Kubernetes manifest repositories: one tab per manifest repo, git awareness, light schema-aware YAML editing, a diff-before-apply Kubernetes workflow with Helm support, and a real embedded terminal built for running Claude Code alongside your manifests.
Status: design phase. The v1 design is complete (DESIGN.md); implementation is tracked in issue #17 against the v1 milestone. There is no usable build yet.
Managing fleets of manifest repos doesn't need an IDE — it needs a small set of things done extremely well:
- Projects — each cloned manifest repo is a top-level tab.
- Git — status badges in the file tree, pull / push, proper diffs. Committing is the terminal agent's job, not m6t's (DESIGN.md §7).
- Editing — CodeMirror-based YAML with Kubernetes schema diagnostics, and markdown preview. Light by design.
- Apply — every cluster mutation goes validate → diff → confirm → apply, with explicit per-project context binding and typed confirmation for protected clusters. Plain YAML and Helm (render→apply or true releases).
- Terminal — real PTY terminals per project (cwd at the repo root), sized for full-screen Claude Code sessions.
Everything m6t does to a repo or a cluster is something you could have typed
yourself (git, kubectl, helm) — it wraps your tooling and credentials,
and never stores either.
Go backend (git/PTY/kubectl/client-go/helm) + Wails system webview frontend (React, xterm.js, CodeMirror 6). Targets macOS, Linux, and Windows. See DESIGN.md for the full architecture, decisions and rationale, and roadmap.
See CONTRIBUTING.md. Issues are organized by area/*
labels; DESIGN.md is the source of truth for scope and
architecture decisions.
Apache 2.0. All dependencies are permissively licensed (MIT/BSD/Apache-2.0), enforced by a CI license gate.