Kimi Family for Github Copilot Chat

Bring the latest Kimi (Moonshot AI) lineup directly into GitHub Copilot Chat, with visible step-by-step thinking and a built-in usage monitor. Kimi Code membership mode exposes k3, k3-256k, kimi-for-coding (now K2.8 Preview under the same ID), and kimi-for-coding-highspeed. Standard API mode exposes kimi-k3, kimi-k2.7-code, kimi-k2.7-code-highspeed, and kimi-k2.6.

Install the Extension from VSCode Marketplace or search Kimi Family for Github Copilot Chat in VSCode Extension Store.

This extension is modeled after glm-for-copilot and adapted for the Kimi (Moonshot AI) OpenAI-compatible API. Huge shout out to @Dootmaan for his foundamental work and generous licensing.

Features

🧩 Seamless Copilot Chat integration — Kimi models show up in the Copilot Chat model picker via the LanguageModelChatProvider API (vendor kimi).
🧠 Visible thinking with effort control — Kimi's reasoning_content streams as Copilot Chat thinking parts. Kimi Code k3, k3-256k, and kimi-for-coding expose Low / High / Max effort and can route to a no-thinking backend when set to Off. kimi-for-coding-highspeed stays always-on. In Standard API mode, kimi-k3 and the K2.7 code family always reason, while kimi-k2.6 also offers Off.
🛠️ Tool calling — full function/tool support (up to 128 tools) for agentic workflows.
🏆 Kimi Code membership support — sign in via OAuth device-code login (browser) or paste a Kimi Code Console API key from kimi.com/code/console. Shared quota with Kimi会员, tracked via the /coding/v1/usages endpoint (weekly quota %, 5-hour window, and Booster/Extra-Usage wallet).
🧭 Automatic mode + region detection — when you paste an API key, the extension checks Kimi Code membership plus both Open Platform regions and updates apiMode / region automatically when it can.
🔁 Confirmed API fallback — if Membership quota is exhausted and you already have a Standard API key with positive balance, the extension offers a one-click switch to Standard API after user confirmation.
💰 Balance monitoring — for the Standard API, a status-bar item and detail panel show your available / voucher / cash balance via GET /v1/users/me/balance. The bar turns red when exhausted.
🔐 Secure credential storage — OAuth tokens, membership API keys, and Standard API keys all live in VS Code SecretStorage (OS keychain), never in settings.json.
🌏 International + Mainland China — platform.kimi.ai (api.moonshot.ai) and platform.kimi.com (api.moonshot.cn) for Standard API; global api.kimi.com for membership.
🔁 Resilient streaming — automatic retries with exponential backoff for HTTP 429 / 5xx.
📝 Token usage reporting — per-request usage (prompt / completion / total / cached) is forwarded to Copilot Chat.

Quick start

Option A: Kimi Code membership (OAuth or API key)

Install the extension.
Run Kimi: OAuth Login for the cleanest membership flow, or run Kimi: Set API Key and choose:
- Sign in with browser (OAuth) — opens a browser for the Kimi device-code login flow (recommended), or
- Paste an API key — paste any Kimi key. Kimi Code keys stay in Membership mode; Open Platform keys are detected automatically.
Pick one of the Kimi Code membership models in the Copilot Chat model picker: k3, k3-256k, kimi-for-coding (K2.8 Preview), or kimi-for-coding-highspeed.

Option B: Standard API (pay-as-you-go)

Install the extension.
Create an API key at platform.kimi.ai/console/api-keys (International) or platform.kimi.com/console/api-keys (Mainland China).
Run Kimi: Set API Key and paste it. The extension auto-detects whether the key belongs to the International or Mainland China platform and updates region for you.
Pick a Standard API model in the Copilot Chat model picker: kimi-k3, kimi-k2.7-code, kimi-k2.7-code-highspeed, or kimi-k2.6.

Automatic switching

After Kimi: Set API Key, the extension probes Kimi Code membership plus both Standard API regions and stores the key in the correct credential slot automatically.
If you are already signed into Membership mode through OAuth or a saved membership key, entering a Standard API key does not force an immediate mode switch. It is saved as a Standard API fallback instead.
When membership quota is exhausted, the usage status bar checks whether a saved Standard API key has positive balance. If it does, the extension prompts you before switching to Standard API and setting the detected region.

Configuration

Setting	Default	Description
`kimi-copilot.apiMode`	`membership`	Active access mode. Usually updated automatically after OAuth login or `Kimi: Set API Key`, but you can still override it manually.
`kimi-copilot.region`	`international`	`international` → `api.moonshot.ai`; `china` → `api.moonshot.cn`. Standard API only. Usually auto-detected after `Kimi: Set API Key`.
`kimi-copilot.baseUrl`	`""`	Override the API base URL. When set, `apiMode` and `region` are ignored.
`kimi-copilot.maxTokens`	`0`	Max output tokens (`0` = API default).
`kimi-copilot.maxRetries`	`3`	Auto-retries for transient failures (429 / 5xx).
`kimi-copilot.thinking`	`enabled`	Legacy compatibility setting. Built-in Kimi models now use the picker-side Thinking Effort control instead.
`kimi-copilot.customModels`	`[]`	Add your own model ids to the picker.
`kimi-copilot.modelIdOverrides`	`{}`	Remap built-in model ids for proxies.
`kimi-copilot.showUsageStatusBar`	`true`	Show the usage status-bar item. In Membership mode, this is also where quota-exhaustion fallback prompts appear.
`kimi-copilot.usageRefreshIntervalMinutes`	`5`	Status-bar refresh interval (1–1440 min). Also controls how quickly exhausted membership quota can trigger a Standard API fallback prompt.
`kimi-copilot.debugLogging`	`false`	Verbose logging to the Kimi output channel.

Commands

Kimi: Set API Key
Kimi: OAuth Login
Kimi: Get API Key
Kimi: Clear API Key
Kimi: Refresh Usage
Kimi: Show Usage Details
Kimi: Open Settings
Kimi: Show Logs

Development

pnpm install
pnpm run compile     # type-check + build to out/
pnpm test            # run vitest
pnpm run lint        # eslint

Press F5 in VS Code to launch an Extension Development Host with the extension loaded.

License

MIT