Add Vachi
as a custom model.

Configure Cursor to use Vachi by adding it as a custom model.

1. Open Cursor Settings (Cmd + Shift + J).
2. Navigate to Models and click Add Model.
3. Enter the model name (e.g., claude-opus-4-8; or use gpt-5.5, or any supported model).
4. Under OpenAI API Key, toggle Override OpenAI Base URL to ON.
5. Enter Base URL: https://gateway.vachiai.com and your VACHI_API_KEY.
6. Select the newly added custom model to start coding.

Codex honors OPENAI_BASE_URL and OPENAI_API_KEY. Pick the path that matches how you launch it.

CLI (one-off, env vars)

export OPENAI_BASE_URL="https://gateway.vachiai.com"
export OPENAI_API_KEY="YOUR_VACHI_API_KEY"
codex

Persistent (CLI + IDE / GUI)

Register Vachi as a provider in ~/.codex/config.toml so the CLI and any IDE harness both pick it up.

model_provider = "vachi"
model = "gpt-5.5"

[model_providers.vachi]
name = "Vachi"
base_url = "https://gateway.vachiai.com"
env_key = "VACHI_API_KEY"

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model. Switch mid-session inside Codex with /model.

Vachi speaks the Anthropic API. Pick the path that matches how you launch Claude Code.

CLI (one-off, env vars)

export ANTHROPIC_BASE_URL="https://gateway.vachiai.com"
export ANTHROPIC_AUTH_TOKEN="YOUR_VACHI_API_KEY"
claude

Persistent (CLI + IDE)

Edit ~/.claude/settings.json so every session and the IDE plugin both pick it up.

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://gateway.vachiai.com",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_VACHI_API_KEY"
  }
}

Desktop app (GUI)

The Claude desktop GUI keeps its settings.json outside your home directory. Open the file at the path for your OS and add the same env block as above.

• macOS: ~/Library/Application Support/Claude/settings.json
• Windows: %APPDATA%\Claude\settings.json
• Linux: ~/.config/Claude/settings.json

In the Cline sidebar, open Settings and configure:

• API Provider: OpenAI Compatible
• Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API key

Cline keeps whatever model you already had selected. Switch on the fly from the chat with /model to pick any supported model.

Pi reads providers from models.json. Add Vachi as an OpenAI-compatible provider:

{
  "providers": {
    "vachi": {
      "type": "openai",
      "base_url": "https://gateway.vachiai.com",
      "api_key": "YOUR_VACHI_API_KEY"
    }
  },
  "models": {
    "claude": { "provider": "vachi", "model": "claude-opus-4-8" }
  }
}

Switch to claude mid-session with Pi's model picker.

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

In Windsurf settings, add a custom provider:

• Open Settings → Cascade → Models.
• Add a custom OpenAI-compatible provider.
• Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API key
• Model: gpt-5.5

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

Roo Code inherits Cline's configuration shape. From the sidebar settings:

• API Provider: OpenAI Compatible
• Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API key
• Model ID: claude-opus-4-8

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

Default Copilot Free / Pro doesn't expose a custom endpoint. For Copilot Enterprise with the bring-your-own-model add-on, register Vachi as the custom OpenAI-compatible provider:

• Endpoint: https://gateway.vachiai.com
• API Key: Your Vachi API key
• Model: gpt-5.5

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model. Individual Copilot plans route exclusively through GitHub's hosted endpoint, so Vachi can't intercept those requests.

Since gpt-5.5 is a custom model, Aider requires metadata about its limits and costs. Create a .aider.model.metadata.json file in your project or home directory:

{
  "openai/gpt-5.5": {
    "max_tokens": 8192,
    "max_input_tokens": 2000000,
    "input_cost_per_token": 0.00000125,
    "output_cost_per_token": 0.00000500,
    "mode": "chat"
  }
}

Then, set the environment variables and run Aider with the openai/ prefix:

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model. Switch mid-session inside Aider with /model openai/<name>.

Configure Continue.dev by adding Vachi as a custom model in your config.json. (Set the provider type to "openai" so Continue knows to use the OpenAI-compatible schema).

{
  "models": [
    {
      "title": "Vachi Gateway",
      "provider": "openai",
      "model": "gpt-5.5",
      "apiKey": "YOUR_VACHI_API_KEY",
      "apiBase": "https://gateway.vachiai.com"
    }
  ]
}

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

Configure Zed to use Vachi as a custom OpenAI endpoint in your settings.json.

"language_models": {
  "openai": {
    "api_url": "https://gateway.vachiai.com",
    "api_key": "YOUR_VACHI_API_KEY"
  }
}

In PearAI, go to the settings and add a custom OpenAI provider.

• API Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Model: claude-opus-4-8

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

In Void's provider settings, choose OpenAI Compatible.

• Endpoint URL: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Model Name: gpt-5.5

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

In your IDE, open CodeGPT settings and configure a custom OpenAI provider.

• Select OpenAI Compatible as the connection type/schema.
• Set the API Base URL to https://gateway.vachiai.com
• Enter your Vachi API Key.
• Provide the Model Name (e.g. claude-opus-4-8; or gpt-5.5, or any supported model).

Register Vachi as a provider in OpenClaw, then add a model under it. The alias is what shows up in the OpenClaw picker.

export VACHI_API_KEY="YOUR_VACHI_API_KEY"

# 1. Register Vachi as a provider pointing at the gateway
oc provider add vachi \
    --api-key "$VACHI_API_KEY" \
    --endpoint "https://gateway.vachiai.com"

# 2. Add a model under it (alias 'vachi' is what shows in the picker)
oc model add vachi claude-opus-4-8 --alias vachi

# 3. Use it
oc model use vachi

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model. You can switch mid-session inside OpenClaw with /model vachi.

Hermes treats any OpenAI- or Anthropic-compatible endpoint as a model backend. In hermes.config.toml, point the [model] block at Vachi:

• provider: openai
• base_url: https://gateway.vachiai.com
• api_key: Your Vachi API key
• model: claude-opus-4-8

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model. See the Hermes provider reference for exact field names if they've shifted between releases.

Register Vachi as a shared model backend in Paperclip's admin dashboard so every agent in the org routes through it:

• Provider type: OpenAI
• Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API key
• Default model: gpt-5.5

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model. Set this at the workspace level to apply to all agents, or per-agent in the Model tab.

In Dify, go to Settings > Model Providers and add a new OpenAI-Compatible provider block.

• Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Model Name: claude-opus-4-8

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

Pass the endpoint URL and API key into your Docker config. Open WebUI will automatically query the /v1/models endpoint to fetch Vachi models.

Update your librechat.yaml custom endpoints configuration to include Vachi as an OpenAI-compatible provider.

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

Go to Settings → LLM Provider. Select Generic OpenAI (Compatible).

• Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Model Name: gpt-5.5

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

In Lobe Chat settings, navigate to Language Models and enable OpenAI.

• API Proxy Address: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Add claude-opus-4-8 to the custom models list.

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

In TypingMind, click on the Model dropdown and select Model Settings.

• Enable Custom OpenAI Endpoint.
• API Endpoint: https://gateway.vachiai.com
• API Key: Your Vachi API Key

In NextChat (ChatGPT Next Web), go to Settings.

• Endpoint: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Add gpt-5.5 to the Custom Models field.

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

In Chatbox settings, select OpenAI Compatible to connect to Vachi.

• API Host: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Add claude-opus-4-8 as a custom model.

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

In the Smart Connections plugin settings in Obsidian:

• Set the API Provider to OpenAI Compatible.
• API Base URL: https://gateway.vachiai.com
• API Key: Your Vachi API Key

Initialize the ChatOpenAI client with Vachi's base URL and API key.

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="gpt-5.5",
    openai_api_key="YOUR_VACHI_API_KEY",
    base_url="https://gateway.vachiai.com"
)

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

Configure the OpenAI LLM class in LlamaIndex to point to Vachi.

from llama_index.llms.openai import OpenAI

llm = OpenAI(
    model="claude-opus-4-8",
    api_key="YOUR_VACHI_API_KEY",
    api_base="https://gateway.vachiai.com"
)

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

Set environment variables for CrewAI to automatically use Vachi as the OpenAI provider.

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

Define the llm_config in your AutoGen setup to point to Vachi's endpoint.

llm_config = {
    "config_list": [{
        "model": "claude-opus-4-8",
        "api_key": "YOUR_VACHI_API_KEY",
        "base_url": "https://gateway.vachiai.com"
    }]
}

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

Add Vachi to your .env file:

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

In n8n, add a new OpenAI connection credential.

• Change the environment variables to override the base URL.
• Or use the HTTP Request node pointing to https://gateway.vachiai.com/chat/completions.

Use Vachi in LangFlow by creating a Custom Component that wraps the ChatOpenAI client.

from langflow import CustomComponent
from langchain_openai import ChatOpenAI

class VachiComponent(CustomComponent):
    display_name = "Vachi Gateway"
    description = "Optimized OpenAI-compatible model."

    def build_config(self):
        return {
            "api_key": {"display_name": "Vachi API Key", "password": True},
            "model_name": {"display_name": "Model",
                          "options": ["gpt-5.5"]}
        }

    def build(self, api_key: str, model_name: str):
        return ChatOpenAI(
            openai_api_key=api_key,
            base_url="https://gateway.vachiai.com",
            model=model_name
        )

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

In your Flowise chatflow, add the ChatOpenAI node.

• Connect Credential: Add your Vachi API Key.
• Base Path: https://gateway.vachiai.com
• Model Name: claude-opus-4-8

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

Go to API Connections (plug icon) → Chat Completion.

• Select OpenAI compatible as the API.
• Enter the Vachi Base URL: https://gateway.vachiai.com under the Reverse Proxy or Custom Endpoint setting.
• Provide your Vachi API Key.

In Agnai settings, configure an OpenAI Custom Endpoint.

• Endpoint URL: https://gateway.vachiai.com
• API Key: Your Vachi API Key
• Set the model to gpt-5.5.

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

Configure a custom API server in Faraday.

• Server Type: OpenAI Compatible
• API URL: https://gateway.vachiai.com
• API Key: Your Vachi API Key

Update your LiteLLM config.yaml to route requests through Vachi.

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.

In the OneAPI admin panel, add a new channel.

• Type: OpenAI
• Base URL: https://gateway.vachiai.com
• Key: Your Vachi API Key
• Models: Type gpt-5.5

The gpt-5.5 above is just an example,claude-opus-4-8 also works. Pick any supported model.

You can add Vachi as a custom model in any workflow, script, or framework. Point your OpenAI SDK client to our endpoint.

from openai import OpenAI

client = OpenAI(
  api_key="YOUR_VACHI_API_KEY",
  base_url="https://gateway.vachiai.com"
)

response = client.chat.completions.create(
  model="claude-opus-4-8",
  messages=[{"role": "user", "content": "Hello!"}]
)

The claude-opus-4-8 above is just an example,gpt-5.5 also works. Pick any supported model.