AI Switcher
Manage and switch named accounts for local AI coding tools from one command.
- Source repo
- burakdede/aisw
- Stars
- ★ 133
- Last updated
- 4d ago
- License
- MIT
- Primary language
- Rust
- FA score
- 58/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificCodex · Claude Code
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- Developers who keep separate work and personal Claude Code logins on one machine.
- Not a fit if
- Teams that do not use Claude Code, Codex CLI, Gemini CLI, or Antigravity CLI
- Source review
- 58/100 · Major gaps
What does this agent do, and when should you use it?
aisw is a local CLI that stores named account profiles for Claude Code, Codex CLI, Gemini CLI, and Antigravity CLI. It can import existing logins or add accounts through OAuth or API keys, then writes credentials to the locations each tool already uses. Switch one tool with a profile or move several tools together through a named context. The repository also offers a desktop app and VS Code and Cursor extensions that use the same profiles. Data lives under ~/.aisw/; secrets go in the OS keyring or files with 0600 permissions, and the README says aisw uses no account, telemetry, or proxy service.
Run aisw init to install the shell hook and import existing logins; add profiles with commands such as aisw add claude work or aisw add codex work-api-key --api-key "$OPENAI_API_KEY". Use aisw use <agent> <profile> for one tool, or aisw context create and aisw context use to map and switch several tools together. Each switch takes a backup first and rolls back if a write fails. aisw status shows active accounts. A folder, repository, or GitHub organization can be bound to a context; the shell hook can warn or block when claude, codex, gemini, or agy starts with a mismatched account. Commands support --json and stable exit codes, including for aisw verify as a CI gate.
- Developers who keep separate work and personal Claude Code logins on one machine.
- Engineers managing multiple Codex CLI, Gemini CLI, or Antigravity CLI accounts.
- Repository contributors who want a work context enforced before launching a coding assistant.
- Users who want to switch several supported coding tools to a matching set of accounts together.
How do you install or deploy this agent?
Install on macOS or Linux with Homebrew or the shell installer; Cargo works on any platform. Windows binaries and checksums are available from Releases.
How do you use this agent?
Install the shell hook and import existing logins, then add profiles and create or activate a context. The API-key example reads its value from an environment variable; adding an API-key profile requires a valid key for the relevant service.
What are this agent's strengths and limitations?
- Manages accounts for four named coding CLIs, with both per-tool switching and grouped contexts.
- Backs up credentials before switching and rolls back when a write fails.
- Can bind folders, repositories, or GitHub organizations to contexts and guard launches with a shell hook.
- Keeps profiles locally and supports OS keyrings or files protected with 0600 permissions.
- Its account management is limited to Claude Code, Codex CLI, Gemini CLI, and Antigravity CLI.
- Setup includes installing aisw and configuring the shell hook; running agent processes retain their original account after a switch.
- OAuth and API-key profiles still depend on accounts or credentials from the relevant services; the README does not state their costs.
- Secure storage varies by platform; live Antigravity credentials use Go's platform storage contract.
How does this agent compare with similar options?
The README identifies signing out and back in, copying auth.json, and editing hidden config as manual alternatives. aisw replaces those steps with named profile switching and takes a backup before writing credentials.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AI Switcher This agent | 58 · Major gaps | CLIFree | ★ 133 | 4d ago | Rust | Codex · Claude Code |
| Clausona Account Manager | 74 · Some gaps | CLI | ★ 54 | today | TypeScript | Codex · Claude Code |
| Obsidian Mind | 79 · Good | CLIFree | ★ 5.4k | 1d ago | TypeScript | Codex · Claude Code |
| CCB Multi-Agent Terminal Workspace | 64 · Some gaps | CLIFree + model costs | ★ 3.6k | 5d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The README says the tool runs locally without telemetry or a proxy and describes keyring or 0600-file storage. This supports limited data exposure, but the privacy and storage claims are mostly project statements and do not show concrete permission boundaries. Switching backs up first and rolls back on write failure, supporting a strong rollback score. The usage flow covers imports, OAuth/API keys, and switching, but the supplied files do not describe confirmation for each sensitive operation. Cargo.toml identifies dependencies and platform keyring backends, but provides no dependency audit or update policy. The installer and shell hook can have external effects whose implementation is not shown here. Publisher identity is unknown; the repository path and username provide attribution, not identity verification.
The README claims transactional switching with rollback on write failure. The CI workflow lists multi-platform builds, formatting, Clippy, tests, and coverage gates; the VS Code adapter test illustrates structured errors and recovery commands. These provide useful static support, but the README's supported-tool list is broader than Cargo.toml's description, and none of these files establishes runtime outcomes, so self-consistency is only partial. Rust dependencies are declared and CI installs Linux system dependencies. Failure messages include error kinds and remediation, though the supplied test example cannot establish coverage of all failure paths.
The README identifies developers managing multiple Claude Code, Codex, Gemini, and Antigravity accounts on one machine, and describes single-agent, cross-agent context, and repository-guard scenarios. It notes that running sessions do not change accounts and lists authentication methods and platform support, which helps bound the capability. Command examples are clear, but activation largely depends on a user running the CLI or installing a shell hook; the exact hook scope is not established by the supplied material. Multi-platform installation and platform-specific credential backends support environment fit.
The README is organized around the problem, features, installation, quick start, supported tools, operation, and documentation. It supplies multiple installation methods, platform coverage, examples, and FAQ/troubleshooting links. Cargo metadata and the MIT license text are explicit. Names and a version appear in Cargo metadata and test samples, but no changelog is included in the supplied files. Project and maintainer attribution comes from the repository path and copyright name; no maintainer responsibility or update path is stated, so those scores remain low.
The tool saves named accounts for multiple coding agents and offers context switching and repository guards, directly addressing account overwrites and using the wrong account in a repository. The README shows common commands and status output; the VS Code adapter test indicates JSON responses can include stable structure, error kinds, and remediation. Benefits are mainly asserted in the README and no time or maintenance savings are quantified, so value and cost-benefit receive moderate scores.
The README makes specific claims about features, installation, storage, and platform support. Cargo.toml supplies checkable version, platform dependency, and license metadata; the CI workflow lists planned checks, and the adapter test samples part of the JSON contract. Together these offer limited cross-source support. CI and test files are static evidence only, not proof that checks ran or passed. The materials do not systematically distinguish verified facts from product claims, so fact/inference separation is partial.
- The supplied repository materials omit the core credential read/write implementation, so this review cannot verify secret handling, permission boundaries, or the transactional rollback claim from source.
- CI workflows and test files show configuration and test examples, not that this revision was built, tested, or covered on every platform.
- README and Cargo.toml describe different supported-tool scopes; check the target release's actual support list before relying on a particular integration.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.3.10 aisw 0.3.10 — Linux binaries run on older distributions
Release notes excerpt
Linux binaries run on older distributions again The aisw-x86_64-unknown-linux-gnu binary in 0.3.8 and 0.3.9 was built against glibc 2.39, so it failed to start on Ubuntu 22.04, Debian 12, and older distributions: ``` /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.38' not found ``` aisw 0.3.10 fixes that. What changed - The x86_64 Linux binary is now built the same way as the aarch64 one, against an older glibc. Any Linux distribution with glibc 2.28 or newer is supported. …View on GitHub ↗ -
v0.3.9 aisw 0.3.9 — Find the right workflow faster
Release notes excerpt
A clearer path from “I need another account” to a safe daily workflow aisw 0.3.9 is a documentation-focused release that makes the project easier to evaluate, adopt, and use correctly. It explains the product around the problems developers actually have: switching between work and personal logins, coordinating different accounts across Claude Code, Codex CLI, Gemini CLI, and Antigravity CLI, and preventing the wrong account from being used in a repository. What is new …View on GitHub ↗ -
Release notes excerpt
Highlights - Hardened auth reliability across Claude Code, Codex CLI, Gemini CLI, and Antigravity CLI. - Claude isolated OAuth switching now stays tied to installs where upstream auth is profile-scoped, with clearer handling for unsupported legacy macOS keychain behavior. - Codex ChatGPT-managed auth is now explicit: durable per-profile isolated CODEX_HOME, bootstrap-only --from-live, and blocked shared-mode switching with typed machine errors and remediation. - Gemini OAuth restores nested live auth/config files more faithfully, reducing drift in managed profiles. …View on GitHub ↗