Continue fails with 401 “Incorrect API key provided” when using VibeProxy as custom OpenAI-compatible endpoint

[Diyago]

I’m trying to use Continue with a local OpenAI-compatible proxy (VibeProxy) running at http://localhost:8317/v1.
The proxy uses ChatGPT OAuth tokens internally, so the external API key is not required and should be ignored.

My config.yaml looks like this:

name: vibeproxy-local
version: "0.1"

models:
  - title: GPT-5
    name: gpt5
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5

  - title: GPT-5 Codex
    name: gpt5-codex
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5-codex

  - title: GPT-5 Codex Mini
    name: gpt5-codex-mini
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5-codex-mini

  - title: GPT-5.1
    name: gpt51
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5.1

  - title: GPT-5.1 Codex
    name: gpt51-codex
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5.1-codex

  - title: GPT-5.1 Codex Mini
    name: gpt51-codex-mini
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5.1-codex-mini

  - title: GPT-5.1 Codex Max
    name: gpt51-codex-max
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5.1-codex-max

  - title: GPT-5.2
    name: gpt52
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5.2

  - title: GPT-5.2 Codex
    name: gpt52-codex
    provider: openai
    model: gpt-4o-mini
    contextLength: 120000
    apiKey: vibe
    baseUrl: http://localhost:8317/v1
    completionParams:
      model: gpt-5.2-codex

tabAutocompleteModel:
  title: Autocomplete GPT-5.2 Codex
  name: gpt52-codex-autocomplete
  provider: openai
  model: gpt-4o-mini
  contextLength: 120000
  apiKey: vibe
  baseUrl: http://localhost:8317/v1
  completionParams:
    model: gpt-5.2-codex

Continue successfully loads the config, but all requests fail with:

401 Unauthorized
Incorrect API key provided: vibe
You can find your API key at https://platform.openai.com/account/api-keys.

The problem:
Continue always sends Authorization: Bearer , and VibeProxy either ignores or replaces it, but Continue’s strict validation seems to reject the response because the API key does not correspond to a real OpenAI API key.

What I need:
Support for using OpenAI-compatible backends that do not require API key validation, such as VibeProxy or other self-hosted proxies. An option to disable API-key validation or allow arbitrary API keys would fix the issue.

Environment:
– Continue v1.2.14
– VS Code
– Base URL: local OpenAI-compatible proxy
– Proxy works with curl and with other clients without API keys

[ranaroussi]

The 401 error with "Incorrect API key provided: vibe" is coming from OpenAI's API, not VibeProxy. This suggests your ChatGPT OAuth session may not be active.

Please verify:

1. Open VibeProxy from the menu bar
2. Check if ChatGPT/Codex shows as Connected (green checkmark)
3. If not, click "Add Account" under Codex to authenticate

VibeProxy replaces the Authorization header with your OAuth token before forwarding requests. If you're not authenticated, it can't do this and the request fails.

Also note: The model names in your config (gpt-5, gpt-5.1, etc.) need to match the actual model names available through your ChatGPT subscription. Try using standard model names like gpt-4o or gpt-4o-mini first to verify the connection works.

Please share:

1. Screenshot of VibeProxy showing your connection status
2. Which providers show as connected?

[ranaroussi]

Friendly reminder: This issue will be closed in 7 days if we don't receive the requested information. Please verify your ChatGPT/Codex connection status in VibeProxy and share a screenshot.

[Diyago]

codex and qwen are connected

Insaf@MacBook-Air-Saryuna ~ % curl http://localhost:8317/v1/models

{"data":[{"created":1762905600,"id":"gpt-5.1-codex","object":"model","owned_by":"openai"},{"created":1762905600,"id":"gpt-5.1-codex-mini","object":"model","owned_by":"openai"},{"created":1763424000,"id":"gpt-5.1-codex-max","object":"model","owned_by":"openai"},{"created":1765440000,"id":"gpt-5.2-codex","object":"model","owned_by":"openai"},{"created":1753228800,"id":"qwen3-coder-plus","object":"model","owned_by":"qwen"},{"created":1754524800,"id":"gpt-5","object":"model","owned_by":"openai"},{"created":1762905600,"id":"gpt-5.1","object":"model","owned_by":"openai"},{"created":1765440000,"id":"gpt-5.2","object":"model","owned_by":"openai"},{"created":1753228800,"id":"qwen3-coder-flash","object":"model","owned_by":"qwen"},{"created":1758672000,"id":"vision-model","object":"model","owned_by":"qwen"},{"created":1757894400,"id":"gpt-5-codex","object":"model","owned_by":"openai"},{"created":1762473600,"id":"gpt-5-codex-mini","object":"model","owned_by":"openai"}],"object":"list"}%  

[Diyago]

I have only the GPT Go plan, which doesn’t provide API access. But I used Qwen and it worked. I thought it would use chat instead of API calls — or not?

from openai import OpenAI

client = OpenAI(
api_key="dummy",
base_url="http://localhost:8317/v1"
)

resp = client.responses.create(
model="qwen3-coder-plus",
input=[
{"role": "user", "content": "Привет, объясни как работает рекурсия на Python"}
]
)

print(resp.output_text)

[ranaroussi]

Thanks for the screenshot! I see Codex and Qwen are connected.

Important clarification: VibeProxy works by using your existing subscriptions (like ChatGPT Plus, Claude Pro, etc.) that include web/app access. It does NOT use API credits.

However, the GPT Go plan may not include the models you're trying to use. VibeProxy proxies requests through your subscription's OAuth session.

For Continue IDE:
The 401 error suggests Continue is sending the API key to the provider. Try:

1. Use Qwen models (since you confirmed those work)
2. Or upgrade to ChatGPT Plus which includes GPT-4 access

What specific model are you trying to use in Continue?