What we learned running every account at once.
Notes on running several AI coding accounts at once: usage limits, parallel agents, git worktrees, and what one account per machine costs you.
- 6 min
Claude Code, Codex and Antigravity, running side by side
What each agent CLI is good at, where they differ in practice, and what it takes to keep all three running at once on one machine.
Claude CodeCodexAntigravitycomparison - 5 min
What it means to let an agent run unattended
An agent working on its own cannot stop to ask permission, so it does not. What that actually costs, and what is worth insisting on before you allow it.
safetyunattendedreview - 5 min
What to do when Claude Code hits its usage limit
Usage limits are a scheduling problem, not an error. How the windows work, why waiting beats retrying, and how to keep working through a limit.
usage limitsClaude Codeplanning - 6 min
Running two Claude Code accounts on one Mac
Claude Code keeps one signed-in account per machine. Here is why the second sign-in evicts the first, and the four ways people work around it.
Claude CodeaccountsmacOS - 7 min
Running AI coding agents in parallel without them colliding
Three agents in one folder overwrite each other. The fix is one working copy per agent, a split that respects file boundaries, and one branch back.
parallel agentsgitworkflow - 6 min
Git worktrees are the missing piece for AI agents
A worktree gives each agent a real working copy on its own branch, sharing one repository. The mechanics, the pitfalls, and why agents need them.
gitworktreesparallel agents - 4 min
Why every AI coding CLI allows one account per machine
Claude Code, Codex and Antigravity all store a single credential per user. The reason is boring, the consequence is not, and it is unlikely to change.
accountscredentialsdesign