Mstro

Changelog

What’s new in Mstro

The latest ships and fixes. Subscribe via RSS or follow updates in the footer of every page.

  1. April 18, 2026
    • feature

    Marketing site relaunch — SEO, FAQ, security page, changelog

    mstro.app ships on Next.js 15 with server-rendered HTML, per-route metadata, structured data for Google Rich Results, and an AI-crawler allow-list (GPTBot, ClaudeBot, PerplexityBot, and more). New sections: FAQ, security/trust page, changelog, docs, and use-case pages. The authenticated app lives on app.mstro.app with Clerk sessions scoped to .mstro.app.

    • FAQ →
    • Security →
  2. April 17, 2026
    • feature

    PWA update checker with hourly polling

    The web app now checks for new versions every hour and prompts you with a toast when an update is ready. No more stale bundles after a deploy. One tap reloads with the newest build.

  3. April 15, 2026
    • feature
    • improvement

    Git worktree registry + persistent tab worktree selection

    Tabs now remember which git worktree you selected across reconnects. A new worktree registry tracks every worktree per orchestra so board and chat tabs resume exactly where you left them — even across device switches.

    • Git worktrees doc →
  4. April 10, 2026
    • improvement
    • fix

    Board healing + refactored board UI state

    The PM Board now auto-heals when boards get into inconsistent states mid-execution (agent crashes, network blips, partial writes). State management refactored for fewer re-renders and faster board-to-tab switching.

    • PM Board doc →
  5. April 5, 2026
    • improvement

    Quality engine simplification

    Quality checks (lint, complexity, AI review) now run through a single pipeline with deterministic ordering and clearer pass/fail signal in the UI. Fewer false positives on generated code.

  6. March 28, 2026
    • security
    • improvement

    Security Bouncer pattern library update

    Expanded default allow/deny patterns for common dev workflows (git, npm, pnpm, cargo, go, make, docker) so long-running agents auto-approve more safe calls without touching Layer 2. Ambiguous calls still route to Claude Haiku for fast classification.

    • Security Bouncer doc →
Mstro
FAQSecurityBlogPrivacyTermsSign in