LLM API Setup for Cursor, Claude Code, Codex

# pip install anthropic from anthropic import Anthropic client = Anthropic( api_key="sk-YOUR_API_KEY", base_url="https://api.xah.io", ) message = client.messages.create( model="claude-opus-4.6", max_tokens=1024, messages=[ {"role": "user", "content": "Write a fizzbuzz in Python."} ], ) print(message.content)

# pip install anthropic from anthropic import Anthropic client = Anthropic( api_key="sk-YOUR_API_KEY", base_url="https://api.xah.io", ) with client.messages.stream( model="claude-opus-4.6", max_tokens=1024, messages=[ {"role": "user", "content": "Tell me a short story."} ], ) as stream: for text in stream.text_stream: print(text, end="", flush=True) print()

# pip install openai from openai import OpenAI client = OpenAI( api_key="sk-YOUR_API_KEY", base_url="https://api.xah.io/v1", ) response = client.chat.completions.create( model="gpt-5.4", messages=[ {"role": "user", "content": "Hello from CKEY"} ], ) print(response.choices[0].message.content)

# pip install openai from openai import OpenAI client = OpenAI( api_key="sk-YOUR_API_KEY", base_url="https://api.xah.io/v1", ) stream = client.chat.completions.create( model="gpt-5.4", stream=True, messages=[ {"role": "user", "content": "Write a short poem."} ], ) for chunk in stream: delta = chunk.choices[0].delta.content or "" if delta: print(delta, end="", flush=True) print()

// npm install @anthropic-ai/sdk import Anthropic from "@anthropic-ai/sdk"; const client = new Anthropic({ apiKey: "sk-YOUR_API_KEY", baseURL: "https://api.xah.io", }); const message = await client.messages.create({ model: "claude-opus-4.6", max_tokens: 1024, messages: [{ role: "user", content: "Write a fizzbuzz in Python." }], }); console.log(message.content);

// npm install @anthropic-ai/sdk import Anthropic from "@anthropic-ai/sdk"; const client = new Anthropic({ apiKey: "sk-YOUR_API_KEY", baseURL: "https://api.xah.io", }); const stream = client.messages .stream({ model: "claude-opus-4.6", max_tokens: 1024, messages: [{ role: "user", content: "Tell me a short story." }], }) .on("text", (text) => { process.stdout.write(text); }); await stream.finalMessage(); process.stdout.write("\n");

// npm install openai import OpenAI from "openai"; const client = new OpenAI({ apiKey: "sk-YOUR_API_KEY", baseURL: "https://api.xah.io/v1", }); const response = await client.chat.completions.create({ model: "gpt-5.4", messages: [{ role: "user", content: "Hello from CKEY" }], }); console.log(response.choices[0]?.message?.content);

// npm install openai import OpenAI from "openai"; const client = new OpenAI({ apiKey: "sk-YOUR_API_KEY", baseURL: "https://api.xah.io/v1", }); const stream = await client.chat.completions.create({ model: "gpt-5.4", stream: true, messages: [{ role: "user", content: "Write a short poem." }], }); for await (const chunk of stream) { const delta = chunk.choices[0]?.delta?.content ?? ""; if (delta) { process.stdout.write(delta); } } process.stdout.write("\n");

curl https://api.xah.io/v1/messages \ -H "Content-Type: application/json" \ -H "x-api-key: sk-YOUR_API_KEY" \ -H "anthropic-version: 2023-06-01" \ -d '{ "model":"claude-opus-4.6", "max_tokens":1024, "messages":[ {"role":"user","content":"Write a fizzbuzz in Python."} ] }'

curl https://api.xah.io/v1/messages \ -H "Content-Type: application/json" \ -H "x-api-key: sk-YOUR_API_KEY" \ -H "anthropic-version: 2023-06-01" \ --no-buffer \ -d '{ "model":"claude-opus-4.6", "max_tokens":1024, "stream":true, "messages":[ {"role":"user","content":"Tell me a short story."} ] }'

curl https://api.xah.io/v1/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-YOUR_API_KEY" \ -d '{ "model":"gpt-5.4", "messages":[ {"role":"user","content":"Hello from CKEY"} ] }'

curl https://api.xah.io/v1/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-YOUR_API_KEY" \ --no-buffer \ -d '{ "model":"gpt-5.4", "stream":true, "messages":[ {"role":"user","content":"Write a short poem."} ] }'

curl "https://api.xah.io/v1beta/models/gemini-2.5-flash-lite:generateContent" \ -H "Content-Type: application/json" \ -H "x-goog-api-key: sk-YOUR_API_KEY" \ -d '{ "contents":[ {"role":"user","parts":[{"text":"Write a fizzbuzz in Python."}]} ] }'

curl https://api.xah.io/api/chat \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-YOUR_API_KEY" \ -d '{ "model":"qwen3.8-27b", "stream":false, "messages":[ {"role":"user","content":"Hello from CKEY"} ] }'

curl https://api.xah.io/api/generate \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-YOUR_API_KEY" \ -d '{ "model":"qwen3.8-27b", "prompt":"Hello from CKEY", "stream":false }'

OpenAI-compatible chat/completions /v1

Cursor

Cursor should use the OpenAI-compatible path with a custom API key and custom base URL.

Required values

Provider: OpenAI-compatible
API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io/v1
Model: gpt-5.4
Route: chat/completions
Plan requirement: Requires a Cursor plan that supports custom API endpoints.

Recommended steps

1Open Cursor Settings and go to AI / Models / API Keys.
2Enable the custom OpenAI key or custom provider option for your build.
3Paste the API key and base URL shown below.
4Choose the suggested model or another currently active model.

Ready-to-copy block

Provider: OpenAI-compatible
API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io/v1
Model: gpt-5.4
Route: chat/completions

Recommended models

gpt-5.4
claude-opus-4.6
claude-opus-4.7
claude-opus-4.8
claude-opus-5.5
claude-sonnet-4.6

Notes

Cursor uses the OpenAI-compatible flow, so the base URL must include `/v1`.
If a model fails inside Cursor, switch to another active OpenAI-compatible model.
After saving, send a short prompt to confirm `chat/completions` works.

Anthropic-compatible messages root URL

Claude Code

Claude Code should be configured through Anthropic environment variables, but the base URL must stay on the root domain.

Install Claude Code

macOS / Linux / Windows (Node 18+): npm install -g @anthropic-ai/claude-code
After install, run `claude` to verify. Settings are saved at `~/.claude/settings.json`.
Linux / macOS: `~/.claude/settings.json`
Windows: `%USERPROFILE%\\.claude\\settings.json`

Required values

API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io
Model: claude-opus-4.6
Route: messages
Max tokens: 1024

Notes

`ANTHROPIC_BASE_URL` should not include `/v1`.
Prefer a model that supports the `messages` route.
If Claude Code asks you to authorize first, finish that login once and relaunch with these env vars.
Requests sent through the `messages` route must include the `max_tokens` field.

Temporary shell session

export ANTHROPIC_BASE_URL="https://api.xah.io"
export ANTHROPIC_AUTH_TOKEN="sk-YOUR_API_KEY"
claude --bare

Temporary PowerShell session

$env:ANTHROPIC_BASE_URL = "https://api.xah.io"
$env:ANTHROPIC_AUTH_TOKEN = "sk-YOUR_API_KEY"
claude --bare

Persistent settings

{
    "env": {
        "ANTHROPIC_BASE_URL": "https://api.xah.io",
        "ANTHROPIC_AUTH_TOKEN": "sk-YOUR_API_KEY",
        "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4.6",
        "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4.6",
        "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-haiku-4.5"
    }
}

Recommended models

claude-opus-4.6
claude-opus-4.7
claude-opus-4.8
claude-sonnet-4.6
gpt-5.4
deepseek-v4-flash

OpenAI-compatible VS Code /v1

VS Code Extensions

Most VS Code extensions work best in OpenAI-compatible mode with a custom base URL.

Required values

Provider: OpenAI-compatible
API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io/v1
Model: claude-opus-4.6
Route: messages or chat/completions
Max tokens: 1024

Continue

name: CKEY API
version: 1.0.0
schema: v1
models:
  - name: claude-opus-4.6
    provider: anthropic
    model: claude-opus-4.6
    apiBase: https://api.xah.io/v1
    apiKey: sk-YOUR_API_KEY
    roles:
      - chat
      - edit
      - apply

Cline

Use the OpenAI-compatible provider, the `/v1` base URL, and your API key from Console.

Roo Code

Use the OpenAI-compatible provider, the `/v1` base URL, and your API key from Console.

Kilo Code

Use the OpenAI-compatible provider, the `/v1` base URL, and your API key from Console.

Recommended models

gpt-5.4
claude-opus-4.6
claude-opus-4.7
claude-opus-4.8
claude-opus-5.5
claude-sonnet-4.6

Notes

Use the OpenAI-compatible provider, the `/v1` base URL, and your API key from Console.
Cursor uses the OpenAI-compatible flow, so the base URL must include `/v1`.
After saving, send a short prompt to confirm `chat/completions` works.
Requests sent through the `messages` route must include the `max_tokens` field.

OpenAI-compatible ~/.codex /v1

Codex CLI

Codex extension for VS Code supports OpenAI-compatible API with custom base URL and API key.

Install Codex CLI

macOS / Linux / Windows (Node 18+): npm install -g @openai/codex
After install, run `codex` to verify. Codex reads auth from `~/.codex/auth.json` and config from `~/.codex/config.toml`.
Config: `~/.codex/config.toml` (Windows: `%USERPROFILE%\\.codex\\config.toml`)
Auth: `~/.codex/auth.json` (Windows: `%USERPROFILE%\\.codex\\auth.json`)

Required values

Provider: OpenAI-compatible
API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io/v1
Model: gpt-5.4
Route: chat/completions

Recommended steps

1Install the Codex extension from VS Code Marketplace.
2Open VS Code Settings and search for Codex configuration.
3Enter your API Key and Base URL in the Codex settings.
4Select a model from the list of active models.

`~/.codex/config.toml` content

model = "gpt-5.4"
model_reasoning_effort = "xhigh"
personality = "pragmatic"

openai_base_url = "https://api.xah.io/v1"

`~/.codex/auth.json` content

{
    "auth_mode": "apikey",
    "OPENAI_API_KEY": "sk-YOUR_API_KEY"
}

Recommended models

gpt-5.4
claude-opus-4.6
claude-opus-4.7
claude-opus-4.8
claude-opus-5.5
claude-sonnet-4.6

Notes

Codex uses OpenAI-compatible endpoints, so the Base URL must include /v1.
After configuration, reload VS Code to apply changes.
Test by opening a code file and trying to chat with Codex.

OpenAI-compatible ~/.factory /v1

Factory Droid

Factory Droid is a CLI assistant that allows custom model configuration via a `settings.json` file.

Install Droid CLI

macOS / Linux: curl -fsSL https://app.factory.ai/cli | sh
Windows (PowerShell): iwr -useb https://app.factory.ai/cli/install.ps1 | iex
After install, run `droid` to verify.
Linux / macOS: `~/.factory/settings.json`
Windows: `%USERPROFILE%\\.factory\\settings.json`

Required values

Provider: openai
API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io/v1
Model: gpt-5.4
Route: chat/completions

Notes

Each entry in `customModels` needs a unique `id` and an incrementing `index` (0, 1, 2...).
Set `provider` to `openai` so Droid uses the OpenAI-compatible flow.
You can add more models by duplicating the object inside the array and changing `model` + `id`.

`~/.factory/settings.json` content

{
    "customModels": [
        {
            "model": "gpt-5.4",
            "id": "custom:ckey-0",
            "index": 0,
            "baseUrl": "https://api.xah.io/v1",
            "apiKey": "sk-YOUR_API_KEY",
            "displayName": "gpt-5.4",
            "maxOutputTokens": 131072,
            "noImageSupport": false,
            "provider": "openai"
        }
    ]
}

Recommended models

gpt-5.4
claude-opus-4.6
claude-opus-4.7
claude-opus-4.8
claude-opus-5.5
claude-sonnet-4.6

Anthropic-compatible JSONC /v1

OpenCode

OpenCode can point its Anthropic provider to your endpoint and use the same API key.

Install OpenCode

Install via npm: npm install -g opencode-ai
After install, run `opencode` to verify. Config lives at `~/.config/opencode/opencode.json` (or `.jsonc`).

Config path

Windows: `%USERPROFILE%\\.config\\opencode\\opencode.json`
Linux / macOS: `~/.config/opencode/opencode.json`

Required values

API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io/v1
Model: claude-opus-4.6
Route: messages
Max tokens: 1024

Ready-to-copy block

{
    "$schema": "https://opencode.ai/config.json",
    "provider": {
        "ckey": {
            "npm": "@ai-sdk/openai-compatible",
            "options": {
                "baseURL": "https://api.xah.io/v1",
                "apiKey": "sk-YOUR_API_KEY"
            },
            "models": {
                "claude-opus-4.6": {
                    "name": "claude-opus-4.6"
                },
                "claude-opus-4.7": {
                    "name": "claude-opus-4.7"
                },
                "claude-opus-4.8": {
                    "name": "claude-opus-4.8"
                },
                "claude-sonnet-4.6": {
                    "name": "claude-sonnet-4.6"
                },
                "gpt-5.4": {
                    "name": "gpt-5.4"
                },
                "deepseek-v4-flash": {
                    "name": "deepseek-v4-flash"
                }
            }
        }
    },
    "model": "ckey/claude-opus-4.6",
    "agent": {
        "explorer": {
            "description": "Fast explorer subagent for codebase exploration",
            "mode": "subagent",
            "model": "ckey/claude-opus-4.6"
        }
    }
}

Recommended models

claude-opus-4.6
claude-opus-4.7
claude-opus-4.8
claude-sonnet-4.6
gpt-5.4
deepseek-v4-flash

Notes

OpenCode should point to the endpoint with `/v1`.
Prefer an active Claude model if your workflow expects Anthropic-style tooling.
If you change the model elsewhere in config, keep the same base URL and API key values.
Requests sent through the `messages` route must include the `max_tokens` field.

OpenAI-compatible openclaw.json /v1

OpenClaw

OpenClaw can use an OpenAI-compatible provider through `openclaw.json` with a `ckey` provider alias.

Config path

Project root: `./openclaw.json`
Windows: `%USERPROFILE%\\.openclaw\\openclaw.json`
Linux / macOS: `~/.openclaw/openclaw.json`

Required values

Provider: ckey / OpenAI-compatible
API Key: sk-YOUR_API_KEY
Base URL: https://api.xah.io/v1
Model: ckey/gpt-5.4
Route: chat/completions

Ready-to-copy block

{
    "agents": {
        "defaults": {
            "model": {
                "primary": "ckey/gpt-5.4"
            }
        }
    },
    "models": {
        "providers": {
            "ckey": {
                "baseUrl": "https://api.xah.io/v1",
                "apiKey": "sk-YOUR_API_KEY",
                "api": "openai-completions",
                "models": [
                    {
                        "id": "gpt-5.4",
                        "name": "gpt-5.4"
                    },
                    {
                        "id": "claude-opus-4.6",
                        "name": "claude-opus-4.6"
                    },
                    {
                        "id": "claude-opus-4.7",
                        "name": "claude-opus-4.7"
                    },
                    {
                        "id": "claude-opus-4.8",
                        "name": "claude-opus-4.8"
                    },
                    {
                        "id": "claude-opus-5.5",
                        "name": "claude-opus-5.5"
                    },
                    {
                        "id": "claude-sonnet-4.6",
                        "name": "claude-sonnet-4.6"
                    }
                ]
            }
        }
    }
}

Recommended models

ckey/gpt-5.4
ckey/claude-opus-4.6
ckey/claude-opus-4.7
ckey/claude-opus-4.8
ckey/claude-opus-5.5
ckey/claude-sonnet-4.6

Notes

OpenClaw should use the `baseUrl` with `/v1`.
Keep `api` set to `openai-completions` for the OpenAI-compatible flow.
You can add or remove models from the list, but keep the `ckey/` alias and API key values aligned with the sample.

Configure each client the right way

Before you configure

Base URL rules

SDKs

Anthropic API request

Anthropic API streaming

OpenAI-compatible request

OpenAI-compatible streaming

Anthropic API request

Anthropic API streaming

OpenAI-compatible request

OpenAI-compatible streaming

Anthropic API request

Anthropic API streaming

OpenAI-compatible request

OpenAI-compatible streaming

Gemini (Google) request

Ollama request

Ollama Generate (/api/generate)

Send a local image as base64

Send an image URL

Base URL

Model

API Key

Cursor

Required values

Recommended steps

Ready-to-copy block

Recommended models

Notes

Claude Code

Install Claude Code

Required values

Notes

Temporary shell session

Temporary PowerShell session

Persistent settings

Recommended models

VS Code Extensions

Required values

Continue

Cline

Roo Code

Kilo Code

Recommended models

Notes

Codex CLI

Install Codex CLI

Required values

Recommended steps

`~/.codex/config.toml` content

`~/.codex/auth.json` content

Recommended models

Notes

Factory Droid

Install Droid CLI

Required values

Notes

`~/.factory/settings.json` content

Recommended models

OpenCode

Install OpenCode

Config path

Required values

Ready-to-copy block

Recommended models

Notes

OpenClaw

Config path

Required values

Ready-to-copy block

Recommended models

Notes