What's New in v4.1.0

• Plugin Manager: Toggle CC plugins on/off directly from the dashboard Settings page
• Git Workflow Config: 4 push modes (Auto-detect, Single Branch, Dev to Main/Master, Feature Branch) with global + per-project settings
• Branch Selector: Auto-detect and switch branches from the Agent Monitor before starting tasks
• Agent Execution Mode: Choose API or Subscription (CC) per agent. Default to Subscription when no API key is set
• API Key Input: Enter and persist your Anthropic API key from the Settings page
• Whitelist Git Staging: Only source code files are staged on commit. Runtime data, junk files, and crypto ticker directories are excluded
• Burn Report Model Breakdown: Opus vs Sonnet token usage and cost tracking
• Auto-Diary Save: Session diary saves automatically on task completion
• Context Window Label: Clear context usage display per agent
• Read-Only Task Detection: Commit buttons hidden when no files changed
• Session ID on Completion: Session identifier visible on task completion screen

Install / Update

pip install --upgrade memstack-skill-loader

Stats

• 127 skills (85 free, 42 Pro)
• 7 dashboard pages, 17 MCP tools
• 3-agent system: Manager, Builder, Reviewer

What's New in v4.0.5

• Builder MCP Tools Selector: Control which MCP servers load per project from the dashboard. Reduces Builder token overhead by blocking servers it doesn't need.
• Real-Time Streaming Output: See Builder output as it works instead of waiting for completion.
• Global MCP Defaults: Set blocked servers once in Settings, applied across all projects. Per-project overrides supported.
• Auto-commit defaults to off: Builder no longer commits without explicit opt-in.
• Builder git prohibition: Builder agent is now prohibited from running git commands directly.

Install / Update

pip install --upgrade memstack-skill-loader

Stats

• 127 skills (85 free, 42 Pro)
• 6 dashboard pages, 17 MCP tools
• 3-agent system: Manager, Builder, Reviewer

What's New in v4.0

• Dashboard: 6-page localhost mission control (Overview, Skills Manager, Burn Report, Memory Browser, Agent Monitor, Settings)
• Agent Runner: 3-agent orchestration (Manager/Builder/Reviewer) with per-agent model selection
• 17 MCP tools for skill discovery, memory, and agent management
• 127 skills (85 free + 42 Pro-exclusive) across 10 categories
• Real-time context window monitoring per agent
• Session diary with AI-authored markdown narratives
• Safe git staging (prevents accidental commits of secrets/runtime data)
• Task completion notifications (browser, tab flash, audio)
• Token usage tracking with estimated costs
• Burn Report with time range filtering (Today/This Week/This Month/All Time)
• Headroom proxy integration (~35-40% token savings)
• Per-agent custom naming and role tooltips
• 3-button commit workflow (Commit / Push / Commit & Push)

Install

/plugin marketplace add cwinvestments/memstack
/plugin install memstack@cwinvestments-memstack

Pro Skills

pip install memstack-skill-loader
claude mcp add --scope user memstack-skills -- python -m memstack_skill_loader
activate_license(key="your-key", email="your-email")

Get Pro at https://memstack.pro

Hotfix (4.0.1)

• Fixed: Pro skills now load on restart via license.json (not just env var)

What's New in v3.5.0

6 New Free Skills

• changelog-generator — Auto-generate CHANGELOG.md from git history
• webapp-testing — Playwright browser testing patterns
• hosted-mcp-catalog — Zero-setup hosted MCP server reference guide
• marketplace-submit — Step-by-step skill/plugin marketplace submission guide
• freelancer-toolkit — Invoicing, time tracking, and analytics for freelancers
• mentor — Real-time plain language narration of decisions as you build

Updated Counts

• Free skills: 83
• Pro skills: 29
• Total: 112

Other Changes

• Restored and cleaned .claude/rules/ folder
• kdp-format removed from git tracking (local only)
• Three-repo architecture documented in CLAUDE.md

100 Skills — 77 Free + 23 Pro Exclusive

18 new Pro skills added in this release covering codebase indexing, diagram generation, browser automation, session restore, drift detection, test generation, security scanning, and more.

Fixes

• Human-readable display names in skill search
• Auto-reindex on session start
• Drift detection auto-fix improved

Pro Skills

Pro subscribers get 23 exclusive skills that never move to free tier. Get Pro at https://memstack.pro

Upgrade: git pull && python scripts/index_skills.py

Full details: https://memstack.pro

What's New

Multi-Agent Pro Skill 🔒

Complete workflow for orchestrating multiple Claude Code instances as a coordinated team:

• Manager/Builder/Reviewer pattern with role-specific prompts
• Setup guide for Let Them Talk bridge + Headroom proxy
• Known issues & workarounds from real production usage
• Phase flow: discussion → planning → execution → review

Token Optimization for MCP Skill Loader

• find_skill preview mode — Returns name + description + score by default (~93% reduction: ~7,500 → ~500 tokens per call). Set full=true for full content.
• list_skills compact mode — New compact=true param returns names-only grouped by category
• Updated tool descriptions to guide CC workflow: find_skill → get_skill

Diary Optimization

• Dedup PreCompact entries — Skips if compaction snapshot saved within last 5 minutes
• Prevent autonomous firing — Diary now ONLY fires on explicit user request, never autonomously
• Multi-agent DORMANT — Builder/Reviewer agents skip diary (Manager handles it)
• Gated SessionStart loading — Lists diary filenames instead of full content (~500+ tokens saved per session)

Previous v3.3.x Features (since v3.3.3)

• .memstack-ignore for per-project skill filtering
• manage_skills MCP tool for enable/disable skills from CC
• skill_stats tool for usage analytics with cache hit tracking
• Version check at startup
• License cache duration persistence fix

Stats

• 82 skills total (77 free + 5 Pro-exclusive)
• Pro skills: consolidate, context-db, api-docs, branching, multi-agent
• 10 categories: Core, Security, Deployment, Development, Business, Content, SEO, Marketing, Product, Automation

What's New

Single Repo Model

• One repo for everyone. No private repo or GitHub invite needed.
• Free tier: 77 fully implemented production skills
• Pro tier: 4 additional skills unlocked via license key (consolidate, context-db, api-docs, branching)
• Get your Pro key at memstack.pro

Skills

• All 77 free skills now fully implemented (stubs replaced with production content)
• 4 Pro-exclusive skills gated at the MCP Skill Loader level
• Covers: deployment, security, databases, git workflows, testing, documentation, business, marketing, SEO, automation, and more

MCP Skill Loader

• New MEMSTACK_SKILLS_DIR env var for custom install paths
• Pro skills bundled in loader package, served only with valid license
• Slug-based gating for reliable Pro skill detection

Hooks & Config

• Updated hooks: gitleaks scanning, session context loading, license nudge
• New hooks: post-tool-monitor, pre-commit-secrets, pre-compact, session-context-load
• .gitleaks.toml for secret scanning

Documentation

• New GETTING-STARTED.md with step-by-step setup for Windows and Mac
• New SKILL-REFERENCE.md with full skill catalog
• All docs rewritten for non-technical users
• MemStackTM branding applied throughout

Other

• Dynamic skill count in start-memstack.bat
• VS Code launch guard (skips if code not found)
• skills/kdp-format/ added to .gitignore (private local skill)