GitHub Copilot High CPU Fix: 5 Tested Solutions

Overview

GitHub Copilot occasionally pins one or more CPU cores at 100%, making your editor unresponsive. The problem appears across VS Code, JetBrains IDEs, and Visual Studio — though VS Code users report it most frequently. Symptoms include lag on every keystroke, fan noise, and the editor’s process monitor showing github.copilot or copilot-agent consuming disproportionate resources. This article covers five tested fixes, ordered from quickest to most involved.

What Causes High CPU Usage

Copilot runs a background language server process (copilot-agent) that continuously analyzes your code context to generate suggestions. When this process spikes, the cause is almost always one of four things:

1. Large or generated files. Copilot tries to tokenize the entire open file and surrounding context. Minified JavaScript, large JSON fixtures, or auto-generated code (protobuf outputs, migration dumps) can push the tokenizer into an expensive loop.
2. Rapid inline suggestion cycling. Typing fast in a file with high suggestion confidence causes Copilot to queue multiple completion requests simultaneously. Each request competes for the same language server thread.
3. Extension conflicts. Other IntelliSense providers — particularly TypeScript, Pylance, or Tabnine running alongside Copilot — can trigger redundant parsing passes. The language server re-processes context that another extension already invalidated.
4. Corrupted extension state. After a failed update or crash, the Copilot extension can enter a state where it repeatedly retries failed requests, creating a tight CPU loop.

There is no single error message. The symptom is the process itself. In VS Code, open the Process Explorer:

Help → Open Process Explorer

Look for extensionHost or a copilot-agent child process consuming >50% CPU sustained over 30 seconds. In JetBrains, check Help → Diagnostic Tools → Activity Monitor.

Solution 1: Restart the Copilot Language Server

The fastest fix. This kills the runaway process without losing your session.

VS Code:

Open the Command Palette (Ctrl + Shift + P on Windows/Linux, ⌘ + Shift + P on macOS) and run:

GitHub Copilot: Restart Server

If the command doesn’t appear, the extension may be in a broken state. Disable and re-enable it instead:

JetBrains:

Go to Settings → Plugins → GitHub Copilot, disable and re-enable, then restart the IDE. JetBrains doesn’t expose a “restart server” action directly.

This works because it forces the language server to drop its current context and reinitialize cleanly. If the spike was caused by a corrupted request queue or a stuck completion, the restart clears it.

Solution 2: Exclude Large and Generated Files

If the problem recurs every time you open specific files, Copilot is choking on file size or content complexity.

Add these to your VS Code settings.json:

{
  "github.copilot.enable": {
    "*": true,
    "json": false,
    "jsonc": false,
    "csv": false,
    "plaintext": false,
    "markdown": false,
    "log": false
  }
}

This disables Copilot for file types where suggestions are rarely useful but tokenization is expensive. Adjust the list to match your project — if you work with large SQL migration files or protobuf outputs, add those types.

For project-level exclusions, create .vscode/settings.json in the repo root with the same config. This avoids affecting your other projects.

The detail docs don’t emphasize: Copilot’s context window doesn’t just include the active file. It pulls from other open tabs. Having five large files open simultaneously multiplies the tokenization cost even if you’re only editing a small one. Close tabs you aren’t actively using.

Solution 3: Disable Inline Suggestions, Use Chat Only

If you want Copilot’s chat and code explanation features but the inline ghost-text completions are the CPU problem, you can disable just the inline component:

{
  "github.copilot.inlineSuggest.enable": false
}

This eliminates the most CPU-intensive part of Copilot — the continuous, keystroke-triggered completion cycle — while keeping Copilot Chat fully functional. Chat requests are on-demand and don’t run a persistent background analysis.

When this is the right call: If you primarily use Copilot for chat-based code generation, /explain, or /fix commands, disabling inline suggestions costs you nothing and dramatically reduces background CPU. This is an underrated configuration that many teams should adopt by default, especially on lower-powered machines or when working in large monorepos.

Solution 4: Clear the Extension Cache and Reinstall

Corrupted state after a failed update is more common than you’d expect, especially if VS Code auto-updates extensions while you’re mid-session.

Cache locations by OS:

# macOS
rm -rf ~/.vscode/extensions/github.copilot-*
rm -rf ~/.vscode/extensions/github.copilot-chat-*

# Linux
rm -rf ~/.vscode/extensions/github.copilot-*
rm -rf ~/.vscode/extensions/github.copilot-chat-*

# Windows (PowerShell)
Remove-Item -Recurse "$env:USERPROFILE\.vscode\extensions\github.copilot-*"
Remove-Item -Recurse "$env:USERPROFILE\.vscode\extensions\github.copilot-chat-*"

Solution 5: Check for Extension Conflicts

Multiple AI coding assistants running simultaneously is a common — and often invisible — cause of sustained high CPU. Each one runs its own language server, and they compete for the same file-parsing resources.

Check if any of these are installed alongside Copilot:

• Tabnine — runs a separate completions engine
• Codeium — similar background tokenization
• Amazon Q (formerly CodeWhisperer) — overlapping inline suggestion provider
• IntelliCode — Microsoft’s older AI completion tool

Disable all competing AI extensions and test. If CPU drops, re-enable them one at a time to identify the conflict.

Beyond AI tools, certain language extensions with aggressive IntelliSense can also amplify the problem. The TypeScript language server in particular is known to interact poorly with Copilot in monorepos with multiple tsconfig.json files. If you’re in a TypeScript project, try setting:

{
  "typescript.tsserver.maxTsServerMemory": 3072
}

This caps the TypeScript server’s memory, which indirectly reduces CPU contention with Copilot’s process.

Still Not Working?

If none of the above resolves the issue:

• Check GitHub’s status page at githubstatus.com — server-side issues can cause the local client to retry aggressively, spiking CPU.
• File an issue on the Copilot feedback repo with your VS Code version, Copilot extension version, and the output from Help → Open Process Explorer.
• Capture a CPU profile in VS Code (Developer: Toggle Developer Tools → Performance tab → Record for 30 seconds) and attach it to your issue. This gives the Copilot team the stack trace they need.
• Try VS Code Insiders — some CPU bugs are fixed in pre-release builds before they hit stable.

FAQ