Chinese documentation is available in README.zh-CN.md.
Copilot Cost Guard shows the current GitHub Copilot Chat model cost in the VS Code status bar, helping you notice when Copilot is using a higher-cost model.
Features
Show the current Copilot Chat model cost multiplier or AIC pricing in the status bar.
Show a legacy multiplier anchor as the main tier when known, while keeping equivalent AIC cost indexes in the tooltip.
Update automatically when the selected Copilot Chat model changes.
Warn when Copilot switches to a premium model.
Use status bar colors to make higher-cost models easier to notice.
Cost signal
Icon
Status bar color
0x / free
Free
Default
Low cost
Green
Default
1x
Check
Default
Medium cost
Warning
Warning background
High cost
Alert
Error background
How It Works
Copilot Cost Guard reads VS Code's local global state database and uses the Copilot Chat model cache to infer the selected model and its cost signal. Older Copilot caches may expose a multiplier such as 1x; newer caches may expose AIC pricing such as In: 500 · Out: 3000 AICs/1M tokens.
It watches the database for changes and also polls periodically as a fallback. This extension does not call a GitHub Copilot API and does not access the Copilot webview DOM.
Configuration
Setting
Type
Default
Description
copilotCostGuard.pollInterval
number
3000
Polling interval in milliseconds. Minimum: 1000.
copilotCostGuard.warnOnPremiumModel
boolean
true
Show a warning when switching to a premium model.
copilotCostGuard.aicBaselineModelId
string
gpt-5.2
Baseline model ID used as 1x for the equivalent AIC cost index.
Limitations
Copilot does not provide an official API for reading the currently selected model.
The extension depends on VS Code and Copilot's internal persisted state, which may change across versions.
Some models, such as Auto, may not have a fixed multiplier or fixed AIC price.
AIC billing has separate input, output, cache-read, and for Anthropic/Claude models cache-write prices, so the equivalent cost index is a comparison aid rather than a fixed request multiplier.
