v0.15.4

Changed

• Describe memory_feedback as a rating operation without directing clients to call it automatically after an answer. Retain ranking effects, unchanged saved text, personal and room scoping, and the read-only membership restriction. No handler, schema, annotation, or permission changes.

Documentation

• The local Claude Code command spells the scope flag --scope user, as
  the hosted command above it already did. It read -s user, the same flag
  (claude mcp add --help in Claude Code 2.1.302 lists -s, --scope <scope>),
  so nothing about what it installs changes. The change is in
  scripts/generate-configs.mjs; the README block,
  docs/configs/claude-code-cli.sh and docs/snippets/claude-code.md, which
  mnemoverse.com/docs mirrors, are regenerated from it.

• npm: https://www.npmjs.com/package/@mnemoverse/mcp-memory-server

• MCP Registry: https://registry.modelcontextprotocol.io/v0.1/servers?search=mnemoverse

What's Changed

• Correct README signup source attribution by @izgorodin in #223
• docs(configs): the local Claude Code command spells the scope flag --scope user by @izgorodin in #224
• Clarify memory_feedback description without post-answer instruction by @izgorodin in #225
• release: v0.15.4 by @izgorodin in #226

Full Changelog: v0.15.3...v0.15.4