How to Manage Multiple AI Coding Agent Accounts on One Machine

The problem usually starts small. You sign into one account for work, another for a personal project, then add a client account. After that, each coding agent stores its credentials and local state differently. Logging out and back in works once, but it is a poor system for a week with several repositories and several accounts.

Use profiles for individual accounts

A profile is one saved account for one tool. Give it a stable name that describes the boundary it represents, such as work, personal, orclient-acme. The account remains in the provider's normal credential location; aisw stores the profile metadata locally and applies the selected profile when you switch.

aisw init
aisw add claude work --from-live
aisw add codex personal
aisw use claude work
aisw status

This is the right level when you are changing one tool at a time. It also gives the desktop app a clear list of named accounts instead of an opaque login state.

Use contexts when the whole setup changes

A context is a named bundle of profiles. It is useful when a client or project needs a matching identity across more than one coding agent, even when each provider uses a different profile name.

aisw context create client-acme \
  --claude acme-claude \
  --codex acme-codex \
  --gemini acme-gemini \
  --antigravity acme-antigravity
aisw context use client-acme
aisw status --json

The important distinction is operational: profiles answer “which account for this tool?” Contexts answer “which complete account setup should be active now?” Use a profile for a single switch and a context for a deliberate work, personal, or client transition.

Bind repositories to the account boundary

Switching accounts manually is still easy to forget. A workspace binding records the context a repository expects, and the shell hook can warn or block when the active setup does not match it.

cd ~/src/client-acme-app
aisw workspace bind . --context client-acme
aisw workspace guard --mode strict

Strict mode is the useful choice once the mapping is trusted. It turns an account mistake into an explicit stop before a coding agent starts in the wrong repository. Start with warn mode if you are auditing an existing machine and want to see mismatches first.

Choose the interface that fits the check

The CLI is the fastest path for scripting, shell hooks, and repeatable setup. AI Switcher Desktop is useful when you want to see active profiles, contexts, repository bindings, and diagnostics before changing them. Both interfaces use the same local state and the same switching model.

Keep the boundary local and verifiable

AI Switcher does not proxy model traffic or create a cloud account. It works with the credential locations and operating-system keyrings already used by the supported tools. After a switch, use aisw status to inspect the recorded state andaisw verify when you need a comparison with the live provider state.

Start with the tool-specific instructions for Claude Code,Codex CLI,Gemini CLI, or Antigravity CLI. For a multi-client setup, read the guide to managing client accounts and then bind the repositories that need a hard boundary.