/ stack

The toolchain I actually run.

The toolchain I run day-to-day. Nothing aspirational — only what's actively in use.

CLAUDE CODE — PLUGINS & MCPS

PERSONAL TIER LIST · MAY 2026

SMUST HAVES
SuperpowersSKILL SUITE
WORKFLOW DISCIPLINE

TDD, planning, debugging, code-review, worktrees — structured dev lifecycle baked in.

claude.com/plugins·obra/superpowers
Lean ContextMCP
TOKEN EFFICIENCY

46 tools, 10 read modes, compression patterns. Cuts context costs dramatically.

leanctx.com·yvgude/lean-ctx
Second BrainOBSIDIAN + PLUGIN
MEMORY LAYER

Obsidian vault + my own SuperBrain plugin. Auto-captures sessions, distills into linked notes, recalls across projects.

obsidian.md·m3talux/superbrain
ANICE TO HAVE
Code ReviewSKILL
QUALITY GATES

Post-feature review against plan & standards. Catches regressions before merge.

claude.com/plugins
UI/UX Pro Max + Magic MCPSKILL + MCP
FRONTEND GENERATION

21st.dev component library via MCP + Pro Max design system. Pair them together.

nextlevelbuilder/ui-ux-pro-max-skill·21st-dev/magic-mcp
Claude-memMCP
CROSS-SESSION MEMORY

Observation memory per repo. Solid for context recall — not a full knowledge graph.

claude-mem.ai·thedotmack/claude-mem
PlaywrightMCP
BROWSER AUTOMATION

End-to-end UI testing & scraping from Claude. Handy, not always needed.

claude.com/plugins·microsoft/playwright-mcp
BSITUATIONAL
GEO AuditSKILL SUITE
AI SEARCH VISIBILITY

Useful when doing SEO / AI-search work. Niche, but powerful in context.

zubair-trabzada/geo-seo-claude
👀WATCH LIST
Engram MCPCOMING SOON
WEAVIATE MEMORY PRODUCT

Potential S-tier second brain. Persistent, org-scoped, knowledge-graph-backed. GA soon.

weaviate.io/blog/engram-deep-dive

PERSONAL OPINIONS ONLY · NOT AN OFFICIAL WEAVIATE RANKING

Editor

Claude Code
primary editor; agents do most of the typing
JetBrains IDEs
IntelliJ, GoLand, PyCharm, DataGrip — surgical edits, debugging, and SQL work

Prompts I treat as code

skills/ scaffolding
reusable, composable, testable workflows pinned per project
commands/
slash-commands I rebuild often (review, brainstorm, ship)
evals on the way
regression suite for the highest-leverage prompts

Multi-agent orchestration

Opus 4.7 lead
top of the loop — refines specs, takes architectural decisions, dispatches sub-sessions
Sonnet implementers
mechanical execution once the work is well-specified. cheaper, faster, parallel-safe
model-by-task discipline
Opus for ambiguity and design, Sonnet for implementation when refinement and docs are set

Supporting tools

iTerm + zsh
still where the agents live
Linear + Slack
team coordination
Obsidian
writing first drafts before they hit /writing

Where I don't use AI

production secrets
never paste credentials, even into a sandbox
live incident triage
raw logs first, agent assistance second
commit message archaeology
git blame is faster than asking