Fix: "Antigravity Server Crashed Unexpectedly"

When the Antigravity server crashes, it usually takes your entire productivity flow with it. You might see a toast notification or the entire UI might simply freeze. This guide focuses on the underlying Node.js and Electron engine stability.

Common Crash Symptoms

Not all crashes show an error box. Look for these "silent" indicators that the server engine is struggling:

• The Spinning Blue Ring: UI stays interactive, but the agent never returns a response. If the whole window locks up instead of crashing outright, work through the Antigravity not responding fixes.
• Terminal Latency: Typing in the terminal feels "laggy" (indicating the main process is choked).
• High Memory Usage: The Antigravity.exe or Antigravity (Renderer) process exceeds 2GB RAM.

March 2026 Crash Causes (v1.20.5)

If you updated to Antigravity v1.20.5 and started seeing crashes immediately, you are not alone. The March 2026 rollout introduced several new failure modes. Here is what the community has identified so far:

If any of the above match your situation, start by checking your quota dashboard. Persistent quota-related crashes have a dedicated fix path — see the quota issues guide before working through the general steps below.

AES-NI / SIGILL Crash on Older CPUs

There is one crash that no amount of restarting, cache-wiping, or GPU tweaking will fix, because it is a hardware problem, not a software one. If the Antigravity window opens fine but the agent dies instantly in a loop — "Antigravity server crashed unexpectedly. Please restart to fully restore AI features." — and the app never recovers, your CPU is almost certainly the cause.

The agent runs as a separate Go language-server binary, and that binary is compiled with the AES instruction set hard-enabled. On a processor that lacks AES-NI, the kernel kills it the moment it hits an AES instruction with an Illegal instruction (SIGILL). If you launch it from a terminal you will see the tell-tale line:

Step 1: Confirm whether your CPU has AES-NI

Before anything else, check whether the instruction set is actually present (or just switched off). On Linux:

If grep returns nothing (or cpuid shows AES instruction = false), the flag is not reaching the OS. That can mean one of two things.

Step 2: If AES-NI is present but disabled in BIOS (the good outcome)

Many desktops and laptops ship with Intel AES-NI set to Disabled in firmware, even though the silicon supports it. Re-enabling it is a permanent fix:

1. Reboot and enter BIOS/UEFI (usually F2, Del, or Esc).
2. Open the Advanced tab (on some boards you must switch to Advanced Mode first).
3. Find the entry named Intel AES-NI (wording varies by vendor) and set it to Enabled.
4. Save & exit, boot back in, and re-run the check from Step 1 — it should now report AES as available.

Because the app cached its broken state while AES was off, clear the cache once after re-enabling:

Step 3: If the CPU genuinely lacks AES-NI (the hard case)

If there is no BIOS toggle and the flag is truly absent, there is no clean fix today. The binary contains thousands of real AES instructions and does not fall back to a software implementation, so the usual Go escape hatch GODEBUG=cpu.all=off does nothing here. As of the latest March 2026 forum threads, a Google moderator only advised updating to the newest release; no software-AES fallback has shipped. Realistic options:

• Run Antigravity on newer hardware (or a remote/cloud host whose CPU exposes AES-NI) — the reliable path.
• Emulate the missing instructions with Intel SDE, which can intercept the binary and provide AES in software. It works but carries a heavy performance penalty, and some distros' default security policy blocks the ptrace it relies on, so treat it as a last resort.

If your machine is this old, this crash is not a bug in your setup — it is a hard requirement of the shipped binary. Track the open Google AI Developers Forum thread for a future software-AES fallback.

Linux & Ubuntu Crash Fixes

On Linux (including Ubuntu 22.04 / 24.04 and Debian-based distros), the "Antigravity server crashed unexpectedly" error has several causes that are platform-specific and not covered by the standard Windows/macOS advice.

1. Missing Keyring / Secret Service

Antigravity stores session tokens in the system keyring. On minimal Ubuntu installs or headless servers, gnome-keyring or libsecret is missing, which causes a silent crash at startup.

After installing, restart with the --password-store=gnome-libsecret flag or add "password-store": "gnome-libsecret" to your Antigravity argv.json.

2. Sandbox Permission Error (SUID)

Electron-based apps on Linux require a user namespace sandbox. If your kernel disables unprivileged user namespaces (common in hardened Ubuntu setups), the renderer crashes immediately.

3. WSL2 Crashes

Running Antigravity inside WSL2 adds an extra layer of complexity. The most common crash trigger is the server process being killed by the Windows host memory compaction. Set a WSL memory cap in %USERPROFILE%\.wslconfig:

Then restart WSL with wsl --shutdown and relaunch.

4. Linux Full Cache Reset

If the above steps do not resolve the crash, perform a full cache wipe for your Linux install path:

Level 0: The Proper Restart

Before tearing apart your config, ensure you've done a "Hard Exit." Changes to internal settings often don't apply until the main process completely terminates.

1. Quit Properly: Don't just close the window. Use Cmd+Q (Mac) or File > Exit (Windows).
2. The 30-Second Rule: Wait 30 seconds before reopening. This allows the backend ports to clear and zombie processes to die naturally.

Level 1: RAM & System Resources

Antigravity is an AI powerhouse. Unlike standard VS Code, it maintains a local vector database of your code. If you have less than 8GB of total system RAM, crashes are almost inevitable during indexing.

Level 2: GPU & Hardware Acceleration

Antigravity uses your GPU to speed up UI rendering. On certain laptops (especially with dual-GPU setups or older Mac Intel chips), this causes a Renderer Process Hang.

The Fix: Disable Hardware Acceleration

1. Go to Settings (Cmd+, or Ctrl+,).
2. Search for hardware-acceleration.
3. Uncheck "Use hardware acceleration when available".
4. Restart Antigravity completely.

Note: This might make the UI feel slightly less "snappy," but it resolves ~40% of sudden crashes on older hardware.

Level 3: Deep Cache Reset

If the server is crashing because of a corrupted workspace index, you need to wipe the local cache. **This will not delete your code**, just the AI's temporary understanding of it.

macOS Cache Reset:

Windows Cache Reset:

Linux Cache Reset:

Success Probability: Clearing the GPUCache specifically solves "White Screen" crashes that happen during startup.

Prevention: Stability Settings

To avoid future crashes, keep the indexer light and monitoring active:

• Ignored Folders: Ensure your .gitignore includes node_modules and dist. For even better performance, use a .antigravityignore file to exclude large binary folders.
• Check the Console: If the app freezes, go to Help > Toggle Developer Tools, click Console, and look for "Out of Memory" or "GPU Process Lost" errors.
• Update carefully: v1.20.5 introduced several crash regressions (see the March 2026 section above). If you updated recently and crashes started, consider rolling back to v1.20.3 while Google ships a patch.