IdeaWell

Project ideas from Hacker News discussions.

Tmux and Claude Code: Sessions, Send-Keys, and the Ceiling

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Theme 1 – Tmux as a stable base for agentic coding
Many commenters rely on tmux (or similar multiplexers) for its reliability and ease of integrating with Claude‑based agents.

“Tmux never crashes.” – oefrha
“I then open named tmux sessions, perform the ssh login via fingerprint, and then tell claude that he is free to take over that session with a given name.” – littlecranky67
“I am doing tmux + Claude but I wanted a bit more.” – petesergeant

Theme 2 – Herdr offers integrated features but suffers from crashes
Herdr is praised for providing an all‑in‑one agentic environment, yet its instability drives users to fall back on tmux or nest tmux inside Herdr.

“Herdr does it all so much better out of the box when it comes to agentic coding.” – thomas_witt
“My only problem with herdr is its propensity to randomly crash and take out the whole session…. Tmux never crashes.” – oefrha
“I still run sessions and processes I really don’t want to lose inside tmux, nested inside herdr.” – oefrha

Theme 3 – Custom or daemon‑oriented alternatives for better persistence and control
Several users have built or are experimenting with their own multiplexers, daemon‑based agents, or remote‑control setups to avoid the limitations of existing tools.

“I created my own “terminal multiplexer for AI agents” after trying out herdr…. My own tool, Jai, connects to multiple machines, handles multiple agents and enables agent to agent communication.” – michaelanckaert
“Like others here I created my own terminal multiplexer as well…. Daemon that holds all sessions… thin client mode… wasm responsive webui…”. – zackham
“Isn’t this the wrong architecture? Fundamentally, a coding agent should be running as a headless/detached process, a daemon.” – teiferer
“You can just install Claude Code on the server and drive it from Claude Desktop via remote control.” – blfr

🚀 Project Ideas

AgentDaemon: Persistent Headless Agent Runtime

Summary

  • Provides a crash‑resistant daemon that runs Claude Code, Codex, or any agent CLI as a detachable service, eliminating session loss from Herdr‑style crashes.
  • Offers encrypted session state, automatic resume, and a lightweight web/iOS dashboard for monitoring and re‑attaching agents from any device.

Details

Key Value
Target Audience Developers who run multiple AI agents locally or on remote servers and need reliable persistence
Core Feature Headless daemon that assigns each agent a UUID, stores encrypted session snapshots, and provides CLI attach/detach plus crash‑auto‑recovery
Tech Stack Rust (daemon + Tokio), SQLite with SQLCipher encryption, Tauri for cross‑platform web/desktop UI, optional SwiftUI iOS client
Difficulty Medium
Monetization Revenue-ready: Subscription $9/mo per user (teams tier available)

Notes

  • Addresses oefrha’s complaint about Herdr randomly crashing and losing work, and teiferer’s suggestion that agents should run as headless daemons.
  • Provides the open‑source, self‑hosted alternative that avoids VC rug‑pull concerns while giving the same remote‑control convenience praised by blfr.

AgentSandbox: Isolated Execution Environment for AI Agents

Summary

  • Runs each agent in a lightweight, namespaced sandbox (firejail/bubblewrap or gVisor) with per‑agent encrypted secret stores, preventing cross‑agent contamination and host exposure.
  • Integrates transparently with existing tmux/zellij workflows or as a drop‑in wrapper for Claude Code/Codex.

Details

Key Value
Target Audience Security‑conscious developers who run multiple agents and need to isolate secrets, filesystem access, and network usage
Core Feature Launches agents inside isolated namespaces with resource limits, secret injection via per‑agent encrypted file, and optional network/filesystem restrictions
Tech Stack Rust (sandbox manager), libseccomp/seccomp‑bpf, optional firejail/bubblewrap, libsodium for secret encryption
Difficulty Medium‑High
Monetization Hobby (open‑source) – optional enterprise support license

Notes

  • Mirrors petesergeant’s desire to sandbox agents and encrypt secrets using byre/happier, and littlecranky67’s use of the Secure Enclave for SSH keys.
  • Directly addresses the security worries raised by cromka and LelouBil about exposing servers to Claude infrastructure.

AgentMobrix: Unified Terminal Multiplexer for Agent Workflows

Summary

  • A modern terminal multiplexer built on xterm.js/wezterm that combines detachable sessions, rich agent metadata panes, image/video clipboard, and end‑to‑end encrypted web/iOS clients.
  • Eliminates the need to juggle many tmux sessions while giving visual status boards and cross‑device continuity.

Details

Key Value
Target Audience Developers who manage numerous agent sessions and want better UI, persistence, and multimedia support
Core Feature Multiplexer with detach/reattach, per‑pane agent UUID/status display, drag‑and‑drop image/paste, remote clipboard over SSH, and encrypted sync to web/iOS front‑ends
Tech Stack TypeScript/React + xterm.js (frontend), Node.js or Rust WebSocket server (backend), SQLite for session state, Tauri for optional desktop app
Difficulty Medium
Monetization Hobby (open‑source)

Notes

  • Solves mkw5053’s copy‑paste and image/video frustrations and f311a’s complaint about narrow, hard‑to‑read layouts by providing a flexible, readable pane.
  • Aligns with petesergeant’s happier web/iOS integration and yoyohello13’s desire for reader‑mode‑like clarity, while offering the persistence of shpool with richer agent‑centric features.

Read Later