Superlogical terminal multiplexer targets persistent developer sessions
Mitchell Hashimoto’s new startup is building a server-side terminal multiplexer, with funding, timing and customers not disclosed.
By Dominic Okoye · Staff Writer
· 3 min read
HashiCorp co-founder Mitchell Hashimoto has launched Superlogical, a startup building a server-side terminal multiplexer for persistent developer work sessions. The Superlogical terminal multiplexer is being pitched as faster than existing tools, but the company has not disclosed funding, valuation, pricing, customers, release timing or headcount.
The company’s starting point is a mature but still critical layer of developer infrastructure: the command line. Superlogical says the terminal is the shared surface across developers, agents, tools and infrastructure, and argues that modern computing lacks a standard stateful workspace for ongoing tasks that span machines and users.
What is a terminal multiplexer?
A terminal multiplexer lets a developer run and switch among multiple command-line sessions, often keeping jobs alive after a user disconnects from a remote machine. Common examples include tmux, Zellij and GNU Screen, tools that are widely used by engineers managing long-running processes, remote servers and parallel workflows.
In a video posted on X, Hashimoto said existing multiplexers carry performance costs because they duplicate user input and must keep local and remote states aligned. He also argued that pairing those tools with newer terminal emulators such as Kitty, Alacritty or Ghostty does not solve the underlying coordination problem.
What is Superlogical building?
Superlogical plans to move more responsibility to the server while using a smarter client on the user’s machine. Hashimoto described an approach in which the client sends input to the server, similar to an SSH session, while the server owns the durable session state. The user could disconnect and return later, while the client handles rendering and scrolling locally rather than waiting for the network for every display operation.
That design depends on the client being more capable than a basic terminal front end. Hashimoto said the connecting client would need to be a sophisticated, standards-compliant client, pointing to one built on libghostty as an example.
Libghostty comes from Ghostty, the GPU-driven terminal emulator Hashimoto began while still at HashiCorp. Ghostty gained attention among terminal-focused developers for combining performance with a broad feature set, and Hashimoto later separated its core engine into a library that other projects could use.
Superlogical’s bet is that the same engine can support a different model for persistent sessions: less duplication between client and server, more durable context for the work itself, and a cleaner base for tools that need to resume, inspect or share running tasks.
Why does Superlogical matter for AI agents?
Superlogical has not laid out a detailed product roadmap, but its framing points beyond a faster replacement for tmux. The company says work needs its own stateful space for context, data and history, a claim that lines up with the way AI coding agents and automation tools need to observe, continue and recover tasks over time.
Gergely Orosz, author of The Pragmatic Engineer, speculated on X that the team could be building an AI-native agentic operating system above existing operating systems. That remains outside what Superlogical has publicly confirmed. For now, the concrete product is narrower: a new terminal multiplexer architecture from a founder with a recent track record in developer tooling.
This story draws on original reporting from The Register.