Alex Sidebar Alternative: Why Phoenix.vu Is the Best AI Coding Agent for Xcode (2026)

Alex Sidebar got the shape right early: a native macOS companion that hooks into the frontmost Xcode window, downloaded by more than 25,000 developers, it became the reference point other tools got compared to — "the Cursor for Xcode," as more than one reviewer put it.

Then, in September 2025, OpenAI hired Alex's three-person team into its Codex division. The company halted new downloads on October 1, 2025. Existing users keep the app and continue to get support — but no new features are shipping, and the roadmap that would have added what it was missing — build integration, App Store compliance checks, a documented security model — stopped there.

Phoenix.vu starts from the same architectural bet — a companion, not a replacement — and keeps going from where Alex left off.

Neither tool asks you to leave Xcode's editor. Alex triggers off keyboard shortcuts inside the frontmost Xcode window; Phoenix.vu opens as a fixed-width, non-activating panel docked beside it. The difference isn't "companion vs. replacement" here — both got that right. It's what each companion is allowed to see and touch once it's there: Alex's own docs ask for broad folder access (Desktop, Documents, Downloads) to analyze files; Phoenix.vu confines every read and write to the project root, with symlinks resolved before the check.

Alex's error resolution reacts to what Xcode's own compiler already surfaced — click, and it proposes a fix for the error Xcode is already showing. Phoenix.vu's run_build tool goes a step further: it calls xcodebuild directly, resolves the scheme itself, and parses the resulting .xcresult bundle into structured diagnostics — so the agent can trigger and read a build on its own, not just respond to one you already ran.

Code search follows a similar gap. Alex indexes the project with VoyageAI embeddings stored locally for semantic search. Phoenix.vu instead parses the codebase with SwiftSyntax and cross-references it with SourceKit's own semantic analysis — a real dependency graph, not similarity search over vectors.

Alex's feature list — quick actions, inline edits, search, image-to-code — covers what a coding assistant does for any project. None of it is Apple-specific compliance work; that was never on its shipped roadmap. Phoenix.vu ships four tools built directly on Apple's own rulebooks:

It's a local, deterministic scan first; the model only judges evidence it's shown, and citations are checked against what the scanner found. A finding that's already certain never reaches the model at all.

Alex's privacy stance is genuinely good: code embeddings stored locally via VoyageAI, no code collection, an opt-out from third-party training. That's a promise about what the company won't do with your data. Phoenix.vu publishes the mechanism instead — a documented, layered threat model you can read and audit:

Only you (your typed messages) and the app's own system rules can direct the agent. Everything else — file contents, tool output, images, documents, build logs, prior turns — is treated as untrusted data, never as instructions.

• Commands are classified, not trusted. Hard-block / approval / auto-allow. Privilege escalation and force-push never run; everything else needs your yes, with a reason shown.
• Chaining can't hide a second command. Pipes, &&, subshells, even embedded newlines force full approval.
• Secrets never reach a model. .env, private keys, and GoogleService-Info.plist are withheld entirely; anything token-like is masked before it leaves the machine.
• File access stays inside the project. Reads and writes are confined to the project root; Alex's own docs ask for Desktop, Documents, and Downloads access up front.
• Every mutation is a snapshot. Writes are reversible; deletions and build-file changes need approval; a runaway-edit breaker pauses large unsupervised changes.

Phoenix.vu runs outside Apple's App Sandbox, since reading arbitrary Xcode projects and spawning xcodebuild can't work inside one — a documented trade-off, compensated by the controls above.

Alex runs two fixed tiers: Pro at $30/month for 600 chat credits, and Unlimited at $200/month, single-device only. Run out of credits mid-month and you're buying top-ups at $12.50 per 250, none of which roll over from the base plan.

Phoenix.vu uses a straightforward prepaid credit system instead. Add credits whenever you need them and spend them only when you run AI tasks — no monthly cap, no subscription tier to outgrow. You still choose the model behind it: Claude, DeepSeek, Kimi, or any OpenAI-compatible endpoint.

Head-to-head: feature by feature

Alex Sidebar figures reflect its public docs, pricing page, and reporting on the OpenAI acquisition as of the app's last shipped state (sources below).

07 Where Alex Sidebar still wins

Alex Sidebar proved that a native companion beside Xcode — not a replacement for it — is the right shape for Apple developer tooling, and its privacy stance was genuinely ahead of the field. But the three people who built it now work on OpenAI's Codex team, and the app hasn't shipped a new feature since October 2025. Phoenix.vu starts from the same architecture and keeps going: the same non-intrusive companion, plus the App Store compliance tooling, the documented security model, and the direct build integration Alex never got to add. For a Swift developer choosing a daily driver in 2026, the question isn't which sidebar looked better first — it's which one still has someone behind it.

09 FAQ