I love command-line tools and terminal UIs. One of my favorite tools these days is Claude Code.
I run many claude code sessions in different tmux windows all in a single monolithic session. I often end up with over 40 different windows in my session, with at least a handful of claude code sessions (and in future, other terminal agentic coding tools as well) for a variety of projects.
The issue that Agent Deck tries to mitigate is managing multiple AI coding session in the terminal.
Too many terminal tabs. Hard to track what’s running, what’s waiting, what’s done. Switching between projects means hunting through windows.
Agent Deck offers 5 small but powerful value additions:
- See everything at a glance.
- Switch in milliseconds
- Never lose track
- Stay organized
- Zero config switching
Now my monolith tmux session isn’t so monolithic - the highly stateful AI agent sessions are now contained in a separate tmux session.
Other Features
- Conversation forking
- Quick attach MCP servers
- Defined once in a global config
Setup
Pretty straightforward [x] Ported over global MCP config (supabase, linear, context7) to the agent deck global config toml file [x] enabled MCP socket pooling
Bug tracker
- [] When you configure a global configuration for MCP, it requires an
.mcp.json to already be created - [] When you are attached to an agent session, you are unable to use vim mode page scrolling, instead you have to scroll with mouse.