Browse this guide

Overview Before you start Claude Code CLI Codex CLI VS Code MiniMax M3 Verify setup Troubleshooting

Ani Router docs

Connect your AI development tools

Pick a guide, copy the complete setup block, and paste it to get started. Claude Code, Codex, VS Code, and MiniMax-M3 all use your Ani Router key.

~2 minutes

Terminal

Claude Code CLI

Connect through the Anthropic Messages protocol for chat, tool use, and streaming.

Read guide

~2 minutes

Terminal

Codex CLI

Connect to Ani Router through the native Responses API with streaming, reasoning items, and tool calls.

Read guide

~3 minutes

Editor

VS Code

Add Claude or GPT models through the Custom Endpoint in VS Code Chat.

Read guide

~3 minutes

Ani Router

MiniMax M3

Use only your Ani Router key to call MiniMax-M3 from Claude Code or an OpenAI-compatible client.

Read guide

Before you start

There are only three steps: create a key, choose a client, and copy the full setup block. Replace YOUR_KEY with your real key and leave everything else unchanged.

01Create an API keysk-anirouter-••••••••02Confirm the model IDclaude-sonnet-4-6 · gpt-5.4 · MiniMax-M3 03Copy and runChoose a tool below

Protect your key

Never commit API keys to Git. Keep personal config in your home directory, and use ignored local files for project-only secrets.

Available

Claude Code CLI

There are only two steps: replace YOUR_KEY with your Ani Router key, then paste the complete block into a terminal. No Claude subscription is required, and do not append /v1/messages to the base URL.

macOS / Linux / WSL: copy everything

Terminal · macOS / Linux / WSL

# 1) Install Claude Code (macOS / Linux / WSL)
curl -fsSL https://claude.ai/install.sh | bash

# 2) Add the Ani Router configuration
export ANTHROPIC_BASE_URL="https://demo.octopusrotuer.com/api/proxy"
export ANTHROPIC_AUTH_TOKEN="sk-anirouter-YOUR_KEY"
export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1
export CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1

# 3) Launch
claude --version
claude --model claude-sonnet-4-6

Windows: copy everything

PowerShell · Windows

# 1) Install Claude Code (Windows PowerShell)
irm https://claude.ai/install.ps1 | iex

# 2) Add the Ani Router configuration
$env:ANTHROPIC_BASE_URL="https://demo.octopusrotuer.com/api/proxy"
$env:ANTHROPIC_AUTH_TOKEN="sk-anirouter-YOUR_KEY"
$env:CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY="1"
$env:CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS="1"

# 3) Launch
claude --version
claude --model claude-sonnet-4-6

What does success look like?

In Claude Code, run /status. It should show the Ani Router base URL, ANTHROPIC_AUTH_TOKEN, and claude-sonnet-4-6. Then ask it to reply only “connected”; a normal reply confirms the setup.

Optional: save it permanently

Merge these fields into your user-level config so future sessions only need the claude command. If the file already exists, merge instead of overwriting it, and never put the key in project-level config.

~/.claude/settings.json · Windows: %USERPROFILE%\\.claude\\settings.json

{
  "model": "claude-sonnet-4-6",
  "env": {
    "ANTHROPIC_BASE_URL": "https://demo.octopusrotuer.com/api/proxy",
    "ANTHROPIC_AUTH_TOKEN": "sk-anirouter-YOUR_KEY",
    "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY": "1",
    "CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS": "1",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4-6",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-sonnet-4-6"
  }
}

Why pin the model and compatibility mode?

Claude Code defaults change over time. Ani Router currently provides claude-sonnet-4-6 and claude-opus-4-6. Compatibility mode disables experimental beta fields that the gateway does not yet fully forward, improving reliability for coding, tool calls, and streaming.

Ani Router available

Codex CLI

Codex supports custom model providers. Save the configuration once, add your Ani Router key, and use GPT-5.4 through the native Responses API.

1. Save the Codex configuration

Merge this into your user-level configuration. If the file already exists, add or replace only these fields and keep your other settings.

~/.codex/config.toml · Windows: %USERPROFILE%\\.codex\\config.toml

model = "gpt-5.4"
model_provider = "anirouter"

[model_providers.anirouter]
name = "Ani Router"
base_url = "https://demo.octopusrotuer.com/api/proxy"
env_key = "ANIROUTER_API_KEY"
wire_api = "responses"

2. Install, add the key, and launch

Terminal · macOS / Linux / WSL

# 1) Install Codex (macOS / Linux / WSL)
curl -fsSL https://chatgpt.com/codex/install.sh | sh

# 2) Add your Ani Router API key
export ANIROUTER_API_KEY="sk-anirouter-YOUR_KEY"

# 3) Launch
codex --version
codex

PowerShell · Windows

# 1) Install Codex (Windows PowerShell)
irm https://chatgpt.com/codex/install.ps1 | iex

# 2) Add your Ani Router API key
$env:ANIROUTER_API_KEY="sk-anirouter-YOUR_KEY"

# 3) Launch
codex --version
codex

3. Verify the connection

In Codex, run /status. It should show provider: anirouter and model: gpt-5.4. Then ask it to read the README and summarize it in one sentence without editing files. A successful task confirms the setup.

Official Codex custom provider documentation · Responses API streaming events

Available

VS Code

VS Code 1.122+ includes Custom Endpoint support. Paste the configuration once to add both Claude Sonnet 4.6 and GPT-5.4; personal BYOK use does not require a GitHub sign-in or Copilot subscription.

1Press Cmd/Ctrl + Shift + P and run Chat: Manage Language Models.
2Choose Add Models → Custom Endpoint and let VS Code open chatLanguageModels.json.
3Replace the file contents with the complete block below, change only YOUR_KEY, and save.
4Return to the Chat model picker and select Claude Sonnet 4.6 or GPT-5.4. If they do not appear, run Developer: Reload Window.

chatLanguageModels.json · Add both models at once

[
  {
    "name": "Ani Router",
    "vendor": "customendpoint",
    "apiKey": "sk-anirouter-YOUR_KEY",
    "apiType": "messages",
    "models": [
      {
        "id": "claude-sonnet-4-6",
        "name": "Claude Sonnet 4.6",
        "url": "https://demo.octopusrotuer.com/api/proxy/v1/messages",
        "toolCalling": true,
        "streaming": true,
        "maxInputTokens": 180000,
        "maxOutputTokens": 20000
      },
      {
        "id": "gpt-5.4",
        "name": "GPT-5.4",
        "url": "https://demo.octopusrotuer.com/api/proxy/chat/completions",
        "apiType": "chat-completions",
        "toolCalling": true,
        "streaming": true,
        "maxInputTokens": 240000,
        "maxOutputTokens": 32000
      }
    ]
  }
]

Verify the connection

In Ask mode, send “Reply only with: connected.” Then switch to Agent mode and send “Read the first line of the currently open file. Do not modify it.” If both work, chat and tool calling are connected.

Key and feature notes

For easy copy-paste, the example puts the key in a user-level configuration file. Do not commit this file or put the key in a project. When you add models through the UI, VS Code can store the key in system secret storage. BYOK covers Chat and utility tasks, not inline suggestions, semantic search, or embeddings; enterprise policy may also disable BYOK.

VS Code Custom Endpoint documentation · VS Code 1.122 release notes

Ani Router available

MiniMax M3

MiniMax-M3 provides a one-million-token context window. Through Ani Router, you do not need a MiniMax account or key. One Ani Router key works with Claude Code and OpenAI-compatible clients.

Use your Ani Router key

Every example below uses sk-anirouter-YOUR_KEY and stays behind Ani Router team access, limits, filtering, audit, and billing. Do not create or enter a MiniMax provider key.

1. Create an Ani Router key

Create a key in the dashboard and make sure its team can access MiniMax-M3. Replace every YOUR_KEY below with this Ani Router key.

Open Ani Router API keys

2. Verify with one command

Replace YOUR_KEY with your Ani Router key and paste the whole block. A content array containing text confirms the key, team access, and MiniMax-M3 route are working.

Terminal · Ani Router Messages API

# Call MiniMax-M3 through Ani Router
curl "https://demo.octopusrotuer.com/api/proxy/v1/messages" \
  -H "Authorization: Bearer sk-anirouter-YOUR_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-M3",
    "max_tokens": 1000,
    "messages": [
      {"role": "user", "content": "Reply only with: connected"}
    ]
  }'

3. Use MiniMax-M3 in Claude Code

Merge these fields into your user-level configuration, replace only YOUR_KEY, and fully restart Claude Code. Keep the base URL at the Ani Router gateway root; do not append /v1/messages.

~/.claude/settings.json · Windows: %USERPROFILE%\\.claude\\settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://demo.octopusrotuer.com/api/proxy",
    "ANTHROPIC_AUTH_TOKEN": "sk-anirouter-YOUR_KEY",
    "CLAUDE_CODE_AUTO_COMPACT_WINDOW": "1000000",
    "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY": "1",
    "CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS": "1",
    "ANTHROPIC_MODEL": "MiniMax-M3",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "MiniMax-M3",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "MiniMax-M3",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "MiniMax-M3"
  }
}

Verify Claude Code

Run /status and confirm the base URL points to Ani Router and authentication comes from ANTHROPIC_AUTH_TOKEN. Run /model and confirm MiniMax-M3 is active.

Existing OpenAI SDK: use Chat Completions

For an existing OpenAI SDK integration, set the base URL to the Ani Router gateway root, use an Ani Router key, and select MiniMax-M3. Use the curl command below to verify the Chat Completions path first.

Terminal · OpenAI-compatible API

# Call through the Ani Router OpenAI-compatible endpoint
curl "https://demo.octopusrotuer.com/api/proxy/chat/completions" \
  -H "Authorization: Bearer sk-anirouter-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-M3",
    "messages": [
      {"role": "user", "content": "Reply only with: connected"}
    ]
  }'

MiniMax M3 invocation guide · MiniMax OpenAI protocol reference

Verify setup

Verify the key, model, and endpoint with a minimal request before opening the client. This separates gateway issues from client configuration issues.

Verify the Claude endpoint

curl "https://demo.octopusrotuer.com/api/proxy/v1/messages" \
  -H "Authorization: Bearer sk-anirouter-YOUR_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 32,
    "messages": [{"role": "user", "content": "Reply only with: connected"}]
  }'

Verify the Codex Responses endpoint

curl "https://demo.octopusrotuer.com/api/proxy/responses" \
  -H "Authorization: Bearer sk-anirouter-YOUR_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "input": "Reply only with: connected"
  }'

Claude Code: run /status and check the gateway and authentication source.
VS Code: select the new model in Chat and send a short message.
Codex: run /status and confirm the provider is anirouter and the model is gpt-5.4.
MiniMax: run the Anthropic curl in its guide first, then confirm the client model is MiniMax-M3.

Troubleshooting

401 / Unauthorized

Confirm the key starts with sk-anirouter- and is active. Claude uses ANTHROPIC_AUTH_TOKEN, Codex uses ANIROUTER_API_KEY, and VS Code needs the real value in apiKey.

Model not found

For Claude, use a Claude ID from the model list. For Codex, confirm gpt-5.4 is active, its API Format is Responses, and the key team can access it.

Claude 400 / Extra inputs are not permitted

Confirm CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1 is set, then fully exit and restart Claude Code.

Model missing in VS Code

Upgrade to VS Code 1.122+, confirm chatLanguageModels.json is a top-level array, then run Developer: Reload Window. Enterprise policy may also disable BYOK.

Codex 404 / provider error

Use only the Ani Router gateway root as base_url; do not append /responses manually. Make sure the deployed gateway includes the Responses endpoint, set ANIROUTER_API_KEY, and fully restart Codex.

MiniMax-M3 is temporarily unavailable

Check only the Ani Router key: make sure it is active, its team can access MiniMax-M3, and its limit is not exhausted. Do not enter a MiniMax provider key. Retry later or contact an administrator if the upstream is busy or concurrency-limited.

Next step

Create your first API key

Create separate keys for devices or projects, set limits, then copy them into your local config.

Open API keys