Robin

Posted on Feb 21

Continue.dev + Claude Max Ban: Fix in 60 Seconds

#ai #vscode #continuedev #claudemax

Continue.dev's Claude integration stopped working for Claude Max subscription users in January 2026. This is the fix.

What Continue.dev is and why it broke

Continue.dev is an open-source AI coding assistant for VS Code and JetBrains. Unlike Cursor or Cline, Continue is free and fully configurable — you bring your own models.

If you configured Continue to use Claude through your Claude Max subscription credentials, that path is now blocked. Anthropic's January 2026 enforcement restricted automated tool access through consumer subscription OAuth tokens.

Continue.dev is actually one of the easiest tools to fix, because it was designed from the start to work with any provider via config.

Fix: update your config.json

Continue.dev stores its configuration in ~/.continue/config.json. You're changing one section.

Before (broken):

{
  "models": [
    {
      "title": "Claude",
      "provider": "anthropic",
      "model": "claude-opus-4-6",
      "apiKey": "sk-ant-..."
    }
  ]
}

After (working — Option A, direct Anthropic API):

{
  "models": [
    {
      "title": "Claude Opus",
      "provider": "anthropic",
      "model": "claude-opus-4-6",
      "apiKey": "your-anthropic-api-key"
    }
  ]
}

Get your Anthropic API key at console.anthropic.com. Pay per token.

Option B: Smart routing (cheaper for mixed workloads)

Continue.dev supports any OpenAI-compatible provider. This routes each request to the cheapest capable model:

{
  "models": [
    {
      "title": "Komilion Balanced",
      "provider": "openai",
      "model": "neo-mode/balanced",
      "apiKey": "ck_your_key",
      "apiBase": "https://www.komilion.com/api/v1"
    },
    {
      "title": "Komilion Premium",
      "provider": "openai",
      "model": "neo-mode/premium",
      "apiKey": "ck_your_key",
      "apiBase": "https://www.komilion.com/api/v1"
    }
  ]
}

With two models configured, you can switch between them in the Continue sidebar. Use Balanced for most tasks, Premium when you need Opus on a hard problem.

Continue.dev-specific settings worth knowing

Tab autocomplete has its own model config in Continue — separate from the chat model. If you use tab completions, update that too:

{
  "tabAutocompleteModel": {
    "title": "Komilion Frugal",
    "provider": "openai",
    "model": "neo-mode/frugal",
    "apiKey": "ck_your_key",
    "apiBase": "https://www.komilion.com/api/v1"
  }
}

Tab autocomplete fires constantly as you type. neo-mode/frugal (~$0.006/call) keeps those completions cheap while saving the better models for explicit chat requests.

Context providers (the @ commands — @file, @codebase, etc.) use the main chat model. These are fine with Balanced or Premium.

The cost difference for Continue.dev users

Continue users tend to run high call volumes — continuous tab completions plus explicit chat requests.

Usage	Direct Opus	Smart routing
Tab completions (300/day)	~$165/day	~$1.80/day
Chat requests (50/day)	~$27.50/day	~$5.00/day
Total	~$192/day	~$6.80/day

The tab completion difference is particularly dramatic — those short, fast calls are exactly what cheap models handle best.

Verify it's working

After updating config.json, reload the VS Code window (Cmd+Shift+P → "Developer: Reload Window").

Test in the Continue sidebar with a simple question. You should see a response. If you're using Komilion, the actual model used appears in the API response headers.

Getting a Komilion API key

$5 free credits, no card: komilion.com

Your key will be in the dashboard immediately after email verification. It starts with ck_.

Full migration guide covering all tools: komilion.com/cline

DEV Community