MemDock open-sourced as a shared memory and Skills workbench for Claude Code and Codex

On November 9, 2026, a developer posted on V2EX announcing MemDock, described as a desktop workbench that centrally manages the memories and Skills of the coding assistants Claude Code and Codex. In the post, the author explains that two kinds of friction emerge over long use of the two tools: test commands may long since have been replaced, yet new sessions still run on stale rules, and updating a Skill requires digging through directories to recall which tool it was originally installed into. MemDock was built to pull such memories and Skills, scattered across different locations, into a single workbench, and is offered in macOS and Windows builds.

According to the post, MemDock's "Memory Overview" lets users search CLAUDE.md, AGENTS.md and Claude Code's automatic memories globally or per project, as well as browse Codex's read-only memory store. Rule files support side-by-side editing and version rollback, and a project memory graph reveals how a rule is connected and jumps back to its source. For cases where the assistant was corrected in conversation but the rule file was never updated, MemDock extracts candidates worth adding and rules worth revising from recent conversations, on a weekly, daily-scheduled or manual basis. Extractions land first in "Pending Memories", where users review the old content, the new suggestion and the original dialogue before choosing to adopt, adopt with edits or discard, and only adopted items are written back to files. The "Skill Management" section lists Skills installed in Claude Code and Codex and supports copying them across the two tools; MemDock additionally provides bidirectional session conversion between them. As of now, no other accounts of MemDock have been published, so the project's status rests on this single announcement.