No good way to manage the config Claude Code generates across scopes
Resolved 💬 4 comments Opened Mar 25, 2026 by ithiria894 Closed May 5, 2026
Claude Code generates configs the longer you use it — memories, skills, MCP configs, commands, agents, rules, plans. But there's no good way to see what's actually loaded or manage it.
Two core problems
1. You can't see what Claude actually loads
- Each category follows different rules: MCP uses
local > project > userprecedence, agents shadow by name, commands have unsupported same-name conflicts, CLAUDE.md walks parent directories, settings merge by precedence chain. - There is no built-in command to show what's effective in the current session.
/memory listonly shows the current project. To see the full picture across all projects, you have to manuallylseach directory andcateach file.
2. Once you see the problem, no good way to fix it
- Want to move a memory from project to global? Multiple rounds of conversation with Claude to find the encoded-path directory,
mvthe file, updateMEMORY.md. - Want to clean up stale memories? You have to
cateach one to see what's inside, then decide one by one. - I found 3 identical MCP server entries across different projects because I added the same server while
cd'd into different directories. You won't discover these duplicates unless you manually check each location.
Possible directions
/effective— show what Claude has loaded in the current session, per category, with source and precedence info/memory list --all-projects— list memories across all projects in one command/memory move <item> --to global— move an item between scopes in one command- A visual dashboard for spatial config management (move, delete, find duplicates, context budget)
I built claude-code-organizer to address this — it provides a "Show Effective" view that computes per-category effective results, context budget analysis, MCP security scanning, and move/delete with undo. 225 tests (84 unit + 141 E2E), zero dependencies. Happy to contribute if this is something the team wants to explore as a built-in feature.
This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗