Job definition
We score the command-center job: many agents, verification, remotes, and Git-visible ops.
This page is written by the imux team as a product capability review. It is not a lab benchmark and not a commissioned press piece. We score the multi-agent daily-driver job on macOS — concurrent CLIs, native performance, Linked Mode automation, browser proof, and a scriptable control plane — not extension-marketplace popularity.
We score the command-center job: many agents, verification, remotes, and Git-visible ops.
Each dimension is 0–10; overall is the arithmetic mean. Today: imux 9.1, VS Code 5.5, Cursor 5.8.
We do not score extension marketplace size or generic IDE popularity. That is a different contest.
The gap is driven by multi-agent fleets, native performance, Linked Mode automation, browser automation, and a scriptable control plane — strengths of the editor era do not transfer one-to-one to the command-center job.
Native terminal-first multi-agent command center.
Electron IDE — editor-first, terminal secondary.
AI on a VS Code base — still Electron, still editor-first.
| Capability | imux 9.1 | VS Code 5.5 | Cursor 5.8 |
|---|---|---|---|
Multi-agent concurrency Run Claude Code, Codex, Clavue, Grok CLI in parallel panes. | 9.8 | 4.2 | 5.0 |
Native macOS performance Swift/AppKit vs Electron shell tax under multi-agent load. | 9.7 | 5.0 | 4.8 |
Terminal quality (GPU) Ghostty-grade GPU terminals as first-class surfaces. | 9.8 | 5.5 | 5.2 |
Automation · Linked Mode Agent mesh detection, supervisor LLM, socket/icc control plane. | 9.6 | 4.0 | 4.5 |
Agent Chat workbench Dense chat column with approvals, resume, and CLI linkage. | 9.4 | 5.8 | 8.6 |
Files · SCM · branch pick Explorers and Git evidence beside agents, not buried panels. | 9.2 | 8.4 | 8.2 |
Remote SSH workspaces Real remote trees and path-aware ops in the same model. | 9.3 | 7.2 | 6.5 |
Browser automation Embedded browser + scriptable checks next to the shell. | 9.5 | 3.5 | 3.8 |
Supervisor · Mission Control Fleet briefs, loops, and evidence-based verification. | 9.4 | 3.8 | 4.2 |
Scriptable control plane Window → workspace → pane → surface targeting via CLI/socket. | 9.6 | 4.5 | 3.5 |
Editor intelligence (LSP · Tab · Composer) In-editor autocomplete, multi-file Composer, and language services. imux is terminal-first — not a full AI IDE clone. | 4.5 | 8.6 | 9.2 |
VS Code and Cursor still center one editor session. imux treats Claude Code, Codex, Clavue, and Grok CLI as parallel workers in split Ghostty panes — with Agent Chat, files, and SCM on the same surface. The product is scored for fleet work, not single-thread chat.
Long multi-agent sessions punish Electron shells. imux is Swift/AppKit with Ghostty-grade GPU terminals, so the workbench stays responsive while several agents stream, notify, and update Git state.
Settings → Automation detects Clavue, Codex, Claude Code, and Grok CLI readiness, pairs with supervisor LLM configuration, and sits next to a scriptable icc/socket control plane for workspaces, notify, and browser checks.
Editor-first IDEs treat the browser as a side trip. imux embeds preview and automation next to agent terminals so UI checks, screenshots, and console evidence stay in the same workspace model.


VS Code and Cursor remain strong for editor-first AI generation. When you need concurrent CLI agents, files and Git evidence, browser automation, and Linked Mode fleets, imux’s product center is more aligned. This review is editorial product opinion; full tables live on comparisons.
Disclaimer: written by the imux team as product capability editorial. Scores weight the agent-era command-center job and are not a third-party lab benchmark or independent press review. Community quotes on Wall of Love remain verifiable public links.