Copilot Auto-Retry

Automatically retries transient GitHub Copilot failures — rate limits, network errors, and service disruptions — so you don't have to sit and click "Try Again" manually.

The Problem

You kick off a task in Copilot agent mode, switch context to something else, and expect it to be done when you come back. Instead, somewhere in the middle, Copilot hit a rate limit or a network blip:

Sorry, you have exhausted this model's rate limit. Please wait a moment before trying again, or switch to GPT-4.1. Learn More

Now instead of a completed task, you're staring at a half-finished result and a "Try Again" button. You have to context-switch back, click retry, wait, and babysit it to completion — wasting the time you thought you were saving.

This extension fixes that. It automatically retries transient failures with exponential backoff so Copilot keeps working while you're away.

Features

• Automatic retry on rate limits — Detects "exhausted this model's rate limit" errors and retries with exponential backoff
• Network recovery — Monitors connectivity to GitHub/Copilot endpoints and retries when your network comes back online
• Exponential backoff with jitter — Spreads retries over time to avoid hammering the API (2s → 4s → 8s → ...)
• Safety guardrails — Hard rate limits (max 15 retries/minute), cooldown periods, and a kill switch to prevent runaway loops
• Status bar indicator — Shows current state at a glance: idle, waiting, retrying, cooldown, or disabled
• Manual retry command — Trigger a retry immediately with Cmd+Shift+R / Ctrl+Shift+R
• Zero configuration — Works out of the box with sensible defaults. All settings are optional.

Install

VS Code Marketplace

Install from VS Code Marketplace

Or search for "Copilot Auto-Retry" in the VS Code Extensions view (Cmd+Shift+X / Ctrl+Shift+X).

Open VSX (for VS Codium, Gitpod, Theia, etc.)

Install from Open VSX

Manual Install

code --install-extension copilot-auto-retry-0.1.0.vsix

How It Works

1. The extension passively monitors Copilot's health using the VS Code Language Model API
2. When a transient error is detected (rate limit, network issue, model unavailability), a retry cycle begins
3. Each retry focuses the chat panel and submits a follow-up message asking Copilot to continue
4. Retries use exponential backoff: 2 seconds, then 4, then 8, up to your configured maximum
5. If all retries are exhausted, a 60-second cooldown kicks in before monitoring resumes

The extension never reads your prompts or responses — it only checks whether Copilot is accepting requests.

Commands

Open the Command Palette (Cmd+Shift+P / Ctrl+Shift+P) and type:

Settings

All settings are under copilotAutoRetry.* in VS Code Settings.

What Errors Does It Handle?

The extension detects and retries these transient failure types:

It does not retry non-transient errors like authentication failures, expired subscriptions, or content filter blocks — those require user action.

Status Bar

The status bar item (right side) shows the current state:

Click the status bar item to toggle the extension on/off.

Requirements

• VS Code 1.90 or later
• GitHub Copilot extension installed

Privacy

This extension:

• Does not read your prompts, responses, or conversation content
• Does not send any data to external services (beyond VS Code's built-in commands)
• Does not wrap or intercept Copilot's request pipeline
• Health probes use a minimal single-token request that is cancelled immediately (zero quota cost)
• Network probes send HTTP HEAD requests to api.github.com and Copilot endpoints only to check reachability

Contributing

See CONTRIBUTING.md for architecture details, design decisions, and development setup.

License

MIT

_{This extension was AI-generated with Claude Opus 4.6 (Anthropic) under human supervision by Max Mazurok.}