Independent second opinions, natively.
You already have an agent defined for one coding CLI, a plain .claude/agents/*.md file. multi-agent runs that same agent on other model CLIs: grok, codex, claude. Each target executes it natively through its own agent runtime, no API keys and no re-implementation. When independent models converge on a finding you get confidence. When they diverge you get coverage.
The same agent file drives both. One is a deterministic tournament, the other a live pair of extra eyes.
multi-model debate
Rounds are convergence gated: min 2, max 4 by default, and the debate stops the moment nobody changes position. /workflows shows exactly the rounds that executed, nothing invented.
live second opinion
You watch the target model think and run tools in real time, and you can talk to it while it works: ask a question or hand it a directive, it answers and keeps auditing.
core.py never learns about a specific model: it normalizes your agent into a portable form, maps its tools onto a neutral capability vocabulary, and each adapter maps those onto its CLI's own tool names. Tools with no mapping are reported, never silently dropped. Adding a CLI is one adapter file; the core changes only for new features. Every spawned process starts in its own group and dies as a tree, no orphans.
Whatever a model produces is normalized into findings-v1: file, line, severity from critical to low, category, description and a proposed fix. Cross-model findings stay unverified claims until the judge confirms them against the real files, and the scoreboard only counts what survived that check.
A full adversarial run is heavy: many model turns, several rounds, minutes to tens of minutes, real token spend. Each round waits for the slowest model. For quick work use fewer models and rounds, or the single-model /multi-agent. Roadmap: Codex and Grok native wrappers, and an MCP server exposing the arena to any MCP-capable CLI.