Editing Blog/2026-01-03/Claude Management

[[File:Screenshot - Contextual Claude Code.png|thumb|`gum` is pretty neat for terminal niceness. I use it in `concode`]]
Recently, like many others I've found that Opus 4.5 is the most useful model for me. It's accurate, [[Blog/2025-12-04/Coders And Conversationalists Switch Places|can talk about sensitive subjects intelligently]], and its use in the Claude Code agent flow makes it far more usable than the web interface. The only problem with it is that one Claude is not ideal for all the tasks I'd like.

So what I have is a simplistic notion of Claude workspaces and a selector tool `concode`

    ➜  claude-workspaces git:(main) tree
    .
    ├── concode
    ├── context-creator
    │   └── CLAUDE.md
    ├── research
    │   └── CLAUDE.md
    ├── video
    │   └── CLAUDE.md
    └── wiki
        └── CLAUDE.md

Then I just use `concode` (which I use as [[Comma command|`,cc`]]) to select the mode I want to operate in and go wild. As an example, the `wiki` context has my [https://github.com/roshan/rowik-mcp rowik-mcp] active and knows about the template pages on my Wiki so I can paste in a HN or Reddit link and it will turn it into the template use and so on.

The `video` context has access to my `~/Movies` directory and knows how I like videos to be turned into GIFs and so on.

So far I'm enjoying this convenience tool.


{{#seo:|description=Discover how one user leverages multiple Claude workspaces and a selector tool to optimize their AI assistant usage for various tasks, from wiki editing to}}

[[Category:Blog]]
[[Category:AI]]