AIXRouter for Copilot

Use AIXRouter models directly from the GitHub Copilot Chat model picker.

AIXRouter for Copilot does not replace Copilot Chat or add a separate chat UI. It registers AIXRouter as a Copilot language model provider, so you can keep using Copilot Chat, Agent mode, workspace context, instructions, and tools while sending model requests through your own OpenAI-compatible router endpoint.

Features

Adds AIXRouter models to the Copilot Chat model picker.
Uses your own API key, stored in VS Code SecretStorage.
Uses https://api.aixrouter.com as the default gateway Base URL and lets you change it when needed.
Loads models from {baseUrl}/openai/v1/models.
Routes Claude requests to the Anthropic Messages endpoint at {baseUrl}/claude/v1/messages; other chat requests use {baseUrl}/openai/v1/chat/completions.
Supports OpenAI-compatible streaming, tool calls, image input, and reasoning output.
Uses Anthropic-compatible automatic prompt caching for Claude requests. The cache point advances with the growing Copilot conversation and can be disabled in settings.
Enriches model metadata with cost, vendor, multimodal, thinking, and context options for supported AIXRouter and AgileRouter base URLs when enabled.
Uses the cached OpenRouter public catalog only to fill metadata fields absent from AIXRouter. It never overrides platform values or uses bundled model snapshots and name-based capability guesses.

Requirements

VS Code 1.116 or newer.
GitHub Copilot Chat installed and signed in.
An OpenAI-compatible AIXRouter endpoint and API key.

Quick Start

Install the extension.
Run AIXRouter: Set API Key.
Open Copilot Chat and choose a AIXRouter model from the model picker.

The extension uses https://api.aixrouter.com as the default gateway Base URL. To use another compatible gateway, run AIXRouter: Set Base URL.

Common Base URLs:

Provider	Base URL
AIXRouter	`https://api.aixrouter.com`
AgileRouter	`https://api.agilerouter.com`

API Routing

AIXRouter treats aixrouter.baseUrl as the gateway root. With the default https://api.aixrouter.com, model discovery uses https://api.aixrouter.com/openai/v1/models. Claude requests use https://api.aixrouter.com/claude/v1/messages with the Anthropic Messages payload shape. All other chat requests use https://api.aixrouter.com/openai/v1/chat/completions.

Commands

AIXRouter: Set Base URL
AIXRouter: Set API Key
AIXRouter: Clear API Key
AIXRouter: Refresh Models
AIXRouter: Force Refresh Model Metadata (OpenRouter + AIXRouter)
AIXRouter: Open Settings

Settings

Setting	Default	Description
`aixrouter.baseUrl`	`https://api.aixrouter.com`	AIXRouter gateway Base URL. Claude requests use `/claude/v1/messages`; other chat requests use `/openai/v1/chat/completions`.
`aixrouter.models`	`[]`	Optional pinned model list. Leave empty to load from `/openai/v1/models`.
`aixrouter.maxTokens`	`0`	Maximum completion tokens. `0` means provider default.
`aixrouter.temperature`	`null`	Optional temperature.
`aixrouter.reasoningEffort`	`high`	Default reasoning effort for models that expose thinking mode.
`aixrouter.enrichPublicModelMetadata`	`true`	Fill missing cost, multimodal, and context metadata from the public model catalog for AIXRouter and AgileRouter base URLs. It never overrides API-provided route limits or capabilities.
`aixrouter.claudePromptCaching`	`automatic`	Send top-level Anthropic `cache_control` for automatic Claude prompt caching, or set to `disabled`.
`aixrouter.metadata.autoRefresh`	`true`	Refresh the versioned, size-limited OpenRouter catalog cache in the background. It only fills metadata fields missing from AIXRouter.
`aixrouter.metadata.openrouterRefreshHours`	`6`	OpenRouter metadata refresh interval in hours.
`aixrouter.debug`	`false`	Write request diagnostics to the output channel. Prompt text is not logged.

Development

pnpm install
pnpm run compile
pnpm run package

Press F5 in VS Code to launch an Extension Development Host.

License

MIT