Bom Agent does not ship its own LLM. The latest AI engines, Claude and ChatGPT (Codex), are already built in, with no extra bill from Bom.
If your primary CLI errors out, the next one in your priority list takes over — one outage cannot freeze a job.
Worker / Automation / Pack / PackBuilder / Standard — each with isolated tool, memory, and MCP exposure.
Codex runs from a launcher in ~/.bom/runtimes/ so your system Node.js never conflicts.
Claude updates itself; Codex auto-updates through npm latest.
bypassPermissions, dontAsk, and allowedTools patterns cleanly separate per-role capabilities.
Bom never charges for LLM usage. The latest AI engines are already built in, no extra bill.