VS Code chat custom endpoint: any model without Copilot

VS Code Chat normally pushes you toward GitHub Copilot. But the Chat panel also supports custom OpenAI-compatible endpoints, which means you can point it at any model apillms supports — Claude, GPT, Grok, DeepSeek or Kimi — without a Copilot subscription or even a GitHub account.

This guide walks through the setup from the VS Code Chat integration guide at /integrations?app=vscode: add a custom endpoint, save your key once, run a one-line installer, reload the window, and pick a model.

Why bother with a custom endpoint

If you already pay for Copilot, VS Code Chat works fine out of the box. But if you don't have Copilot, or you want to use a model Copilot doesn't offer — a cheaper DeepSeek model for routine edits, or Claude for harder refactors — the Custom Endpoint path lets you wire your own API key in and choose per-conversation which model answers.

Because apillms is pay-as-you-go, you're not locked into one model or one subscription tier. You top up credits once and spend them on whichever model fits the task.

Step 1: Get your API key

Sign up at /login if you haven't already — new accounts get 50 free credits to test with. Every account has exactly one API key, created automatically. Go to the API keys page in the console and click Show key to copy it. There's no "create key" button and no multiple keys; if the key is ever compromised, you roll it instead, which disables the old one.

Step 2: Add a custom endpoint in VS Code

Open the Chat panel in VS Code, then:

1. Click Chat → Manage Models
2. Choose Add Models
3. Select Custom Endpoint
4. Pick Chat Completions as the API type

Paste your apillms API key here and save it. This is the "save the key once" step — after this, VS Code remembers the credential and you won't need to re-enter it.

Step 3: Run the one-line setup command

With the key saved, run the installer for your OS. On macOS or Linux:

curl -fsSL 'https://apillms.com/v1/setup-vscode?os=mac' | bash

On Windows (PowerShell):

irm 'https://apillms.com/v1/setup-vscode?os=windows' | iex

This script configures the endpoint details (base URL and model list) so VS Code Chat knows how to talk to apillms. The OpenAI-compatible base URL behind this is https://apillms.com/v1, the same one used for scripts and SDKs if you ever want to call the API directly.

Step 4: Reload the window and pick a model

Reload the VS Code window (Command Palette → "Developer: Reload Window"). Then open the model picker in Chat. You'll see entries ending in the site name — these are the apillms-backed models. Pick one and start chatting.

Any model id on sale works this way, including:

• claude-sonnet-5 — strong general coding model
• claude-opus-5-5 — for harder reasoning or large refactors
• gpt-6-astra and gpt-5.5 — OpenAI-family options
• deepseek-v4-flash — cheap, fast model for routine edits

Swapping models later is just a matter of reopening the picker — no need to redo the setup command.

What it costs per model

apillms bills per token at the live price shown on /pricing. For reference, here are per-1,000-token chat message costs (1,000 input + 500 output tokens) and the same thing at scale:

Because credits never expire and failed requests aren't billed, there's no reason to over-buy a pack up front. The Starter pack at $5 (1,000 credits) is plenty to try all five models in VS Code Chat before deciding which one fits your workflow.

Checking your usage

Once you're chatting through VS Code, you can track spend in the console. The Overview page shows balance, today's spend and the last 7 days. The Usage page breaks it down by product, with filters by type and status, and an Export CSV option if you want the raw numbers. If you're running both VS Code Chat and something like Claude Code against the same key, this is the easiest way to see which one is actually driving the bill.

Comparing to OpenRouter prices

If you've used OpenRouter with VS Code's custom endpoint feature before, the setup is conceptually the same — paste a key, pick a base URL, choose a model. The difference is pricing: apillms lists input/output/cache prices directly in the console, and you can see a live side-by-side at /compare. For models like deepseek-v4-flash, OpenRouter's listed price is lower per raw token for that specific model, but for most of the Claude and GPT-family models above, apillms comes in lower than OpenRouter's listed rates. Check /compare for the current numbers before deciding.

Takeaway

You don't need Copilot to use VS Code Chat. Save your apillms key once under Custom Endpoint, run the one-line setup script for your OS, reload the window, and pick any model ending in the site name. From there it's normal pay-as-you-go billing, trackable in the console, with the freedom to switch models per task instead of being stuck with one provider.

If you also use Claude Code day to day, the same API key works there too — see Set Up Claude Code with apillms in 5 Minutes for that setup.

FAQ

Do I need a GitHub or Copilot account to use this setup?

No. The Custom Endpoint path in VS Code Chat works independently of GitHub Copilot — you only need your apillms API key.

Can I switch models after the initial setup?

Yes. Once the endpoint is configured, open the model picker in Chat any time and choose a different model ending in the site name — no need to rerun the setup command.

What if the one-line installer fails partway through?

Failed requests aren't billed, but if the installer itself fails, just rerun the command for your OS (mac or Windows) — it's safe to run again.

How do I see how much VS Code Chat is costing me?

Check the Overview and Usage pages in the console; Usage lets you filter by product and export a CSV if you want to isolate VS Code Chat spend from other tools using the same key.