One base URL, the tools you already use.

1. Open the Cline settings panel and set API Provider to OpenAI Compatible.
2. Base URL, API Key and Model ID as below.
3. Under Model Configuration, fill the numbers so Cline budgets correctly rather than guessing.

What we checked. Configuration read from Cline's own OpenAI-Compatible page; the endpoint answers every request shape Cline sends, including tool calls and streamed usage. Cline docs

1. Settings → Models. Add a custom model with the id below.
2. Put your tiyuvta key in the OpenAI API Key field, switch on Override OpenAI Base URL, and set it to the base URL below.
3. Press Verify.
4. Turn OFF the other models in that list. With the override set, Cursor can still route a request meant for another vendor's model through this base URL, and the failure looks like our endpoint rejecting a model it has never heard of.

What we checked. Cursor sends ordinary Chat Completions, which this endpoint serves. The caveat below is Cursor's, not ours, and it is the one that wastes an afternoon. Cursor docs

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "tiyuvta": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "tiyuvta",
      "options": {
        "baseURL": "https://api.tiyuvta.ai/v1",
        "apiKey": "{env:TIYUVTA_API_KEY}"
      },
      "models": {
        "qwen/qwen3.8-27b": { "name": "Qwen3.8 27B" }
      }
    }
  }
}

What we checked. Config shape read from opencode's providers page. The model key must match an id from GET /v1/models, which is why the snippet uses the exact id the catalogue returns. opencode docs

{
  "providers": {
    "tiyuvta": {
      "baseUrl": "https://api.tiyuvta.ai/v1",
      "api": "openai-completions",
      "apiKey": "$TIYUVTA_API_KEY",
      "models": [
        {
          "id": "qwen/qwen3.8-27b",
          "name": "Qwen3.8 27B",
          "reasoning": true,
          "input": ["text", "image"],
          "contextWindow": 262144,
          "maxTokens": 32768,
          "cost": {
            "input": 0.40,
            "output": 2.90,
            "cacheRead": 0.10,
            "cacheWrite": 0
          }
        }
      ]
    }
  }
}

What we checked. Written against Pi's models.json reference. Every compat flag Pi offers for partial OpenAI endpoints was tested here and none is needed: the developer role, reasoning_effort and streamed usage all work, so the defaults are correct. Pi docs

from openai import OpenAI

client = OpenAI(
    base_url="https://api.tiyuvta.ai/v1",
    api_key=os.environ["TIYUVTA_API_KEY"],
)

reply = client.chat.completions.create(
    model="qwen/qwen3.8-27b",
    messages=[{"role": "user", "content": "Explain KV caches in one sentence."}],
    # This model reasons first. A small budget is spent entirely on reasoning and
    # returns empty content with a 200 — give it room.
    max_tokens=1024,
)

What we checked. Two lines change: base URL and key. This is the path every other tool on this page is doing on your behalf. our docs

export OPENAI_API_BASE=https://api.tiyuvta.ai/v1
export OPENAI_API_KEY=$TIYUVTA_API_KEY

aider --model openai/qwen/qwen3.8-27b

What we checked. Aider routes OpenAI-compatible providers through LiteLLM, whose openai/ prefix path was exercised against this endpoint directly. Aider docs