How to manage multiple coding agents from Neovim without living in the terminal
4310sy · x · 2026-08-04
This long blog post describes a workflow for managing multiple coding agents without living in the terminal.
The problem
The author works with agents like Codex and Claude Code while editing in Neovim, but constantly switching to terminals to check status becomes the bottleneck.
The approach
- Keep editing inside Neovim.
- Use Herdr to manage PTYs and running processes so sessions keep going even if the UI closes.
- Build signalbox.nvim to surface only the agents that are actually waiting for human input.
The interaction model
The workflow is split into four steps: notice, inspect, instruct, return.
- Herdr tracks states such as blocked, working, done, idle, and unknown.
- Signalbox polls these states and notifies Neovim when an agent becomes blocked or done.
- A small persistent indicator can stay visible in Neovim while transient notifications disappear.
Why it matters
The post argues that the goal is not to create yet another dashboard, but to let developers stay in their editor, notice only the right moments, and get back to code quickly.
More from coding & agent
- LangGraph Studio Introduces Replay: Swap Variables and Re-run Exact Traces for Agent Debugging — hwchase17 · 2026-08-04
- Local LLM user asks which agent harness is best for lightweight daily tasks — _n1vk · 2026-08-04
- MiniMax H3 ComfyUI test shows better video quality at higher resolution and steps — ayakitodev · 2026-08-04
- GitHub repo backs an AI bug-hunting benchmark with scripts, logs, and caveats — PawelHuryn · 2026-08-04
- Jitsu: An Open-Source, Self-Hosted Alternative to Segment for Data Pipelines — tom_doerr · 2026-08-04
- A simple checklist for verifying AI-generated math and code in economics research — Afinetheorem · 2026-08-04