A git-tracked memory layer for Claude Code

What it does Storybloq scaffolds a .story/ directory inside your repo to hold JSON and markdown files representing tickets, issues, roadmap phases, session handovers, and lessons learned. A CLI and MCP server let Claude Code (or Codex) read and mutate this state directly, while a /story skill loads context at the start of every session. The result is continuity: the agent knows what was built, what’s broken, and what to do next without relying on human recaps.

The interesting bit The project treats AI context as structured, versioned data rather than prompt stuffing. It also federates across multiple repos via an orchestrator node that tracks cross-repository dependencies and blockers, so a web app can know its API ticket is still open.

Key highlights

• 53 MCP tools expose the entire project state to Claude Code without subprocess spawning
• Autonomous mode (/story auto) can drive a ticket through planning, implementation, tests, and review with handovers at each checkpoint
• Federation supports multi-repo projects with dependency graphs and cross-node blockers
• A companion Mac app provides a native sidebar that watches .story/ and updates live
• Everything lives in git except ephemeral snapshots; no external database required

Caveats

• Licensed under PolyForm-NC 1.0, so commercial use is off the table
• Requires Claude Code or Codex CLI 0.130.0+ and Node.js 20+; it is not a generic editor plugin
• The Mac app is a separate product distributed via the App Store

Verdict Worth a look if you are already living in Claude Code and tired of re-explaining your project every morning. Skip it if you need a commercial license, use a different AI client, or prefer lightweight CLAUDE.md files.