Claude usage ↻ refresh

Loading this server's usage…

Session

Where the AI works for THIS chat (e.g. /opt/preloved.ph). Blank = the server's own directory. Applies on your next message.
Applies on your next message.
Split shows the AI's file edits & background tasks in their own column.
Scales the message thread and actions text on this device (independent of browser zoom). Smaller = more on screen.

Skills

Pulls only the skills assigned to this server (Manage servers → Skills). Assigning already auto-syncs — this is a manual refresh.

Pick a server on the left, or add one with ＋. Sessions keep running even after you close this tab.

Actions & background tasks
Drop files to attach
Plan
Files
↓
Drop to upload

        
        

        
      

    


    

      

        Terminal
        

          
          
          
          
          
          
        

      

      

    

  


    

      

        Preview
        
        

          
          
          
          
          
          
          
          
          
        

      

      

        
        

          
Console
            
            
          

          

        

      

      
Loads from your browser, so the address must be reachable from this device. A site that sends X-Frame-Options or CSP frame-ancestors will refuse to appear here — use open for those.

    


  

    

      
Rewind this chat

      
This message and everything after it are deleted, and the agent's session transcript is truncated so it genuinely forgets them — not just hidden from view. The chat keeps its history up to that point.

      
      

      

        
        
      

    

  


  

    

      
Terminal text · select and copy normally

      
xterm draws the terminal to a canvas, so the phone's selection handles and Select All never line up with it. This is the same scrollback as ordinary text — tap and hold, drag the handles, or Select All.

      
      

        
        
      

    

  


  

    

      
Skills library · shared definitions

      
This is the shared library of skill definitions (lives on the hub, no GitLab). Defining a skill here does not put it on any server. To give a server a skill, go to ＋ Manage servers → (server) → Skills and tick it — each server only gets the skills you assign it (so a Japanese skill never lands on a Tagalog box).

      
      

      

        

        

      

      
    

  


  
  

    

      
Use · skills and modules for this chat

      
Tick what this task needs. Skills are how-to guides the AI follows; modules are proven subsystems it reads before building that part. One search covers both. The choice rides along with your next message and then clears, like an attachment.

      

        
        
        
      

      
      

      

        
        
      

    

  


  

    

      
Modules · reference catalog

      
Proven implementations of web-app subsystems, extracted read-only from collabz.ph, runway.ph, pribado, preloved.ph and markthedate. Unlike Skills there is no per-server assignment — every server gets the whole catalog automatically, and the agent is told where it lives. It reads a module before building that subsystem from scratch.

      
One module = one feature, each a full slice: schema, service, API routes and admin screen. Source of truth: thokzz/konzoles-modules

      
      

      

        
        
      

      

      
    

  


  

    

      
MCP library · shared connector definitions

      
MCP servers give the AI extra tools (APIs, databases, services). This is the shared library of connector definitions — defining one here does not enable it anywhere. To give a server a connector, go to ＋ Manage servers → (server) → Connectors and tick it. Each server only gets the connectors you assign it.

      

      

        
      

      
    

  


  

    

      
      
Sign in to your control plane

      

      

    

  


  

    

      
Manage servers

      
Enroll a new server, or remove an existing one below.

      
      

      

      
      
      

      

        
        

        
        
After it connects, set its Claude login: run claude setup-token and put CLAUDE_CODE_OAUTH_TOKEN in /etc/konzoles/secrets.env, then restart the service.

      

      

    

  


  

    

      
New chat

      
Name it and pin the directory the AI will work in for this chat.

      
      
      
      
Each chat targets one app. Type /opt/ and pick the app from the list — the AI works there and won't guess. (Change it later in the chat's settings.)

      

    

  


  

    

      
⑂ New subchat

      
A nested child conversation on its own model, inheriting this chat's working directory. Pick the model to fit the task — Haiku for grunt work, Sonnet/Opus for heavier reasoning.

      
      
      
      
Fill in a task → it runs as a one-shot and auto-posts its result back here. Leave it empty → you get a quiet nested chat you drive yourself (push a result up anytime with Post ↑ parent).

      

    

  


  

    

      
Guide — servers & the AVECO AI box