Skip to content
FluxCDNFluxCDN
esc
On this page

DocsTools guide

FluxCDN AI in your tools

One address and one key for all 581 models. Keep the tools you work with: coding agents, editors, chat apps, automation and your own code.

What you need

OpenAI-compatible address

https://ai.fluxcdn.cloud/api/v1

For most tools and every OpenAI SDK. Codex uses it too.

Anthropic-compatible address

https://ai.fluxcdn.cloud/api

For Claude Code and the Anthropic SDKs. They add /v1/messages themselves.

API key

flux_live_…Get an API key

Create one on the API keys page. It is shown in full only once, so copy it right away.

Both addresses take every model. Copy the model id from the models page, such as anthropic/claude-sonnet-5.5 or deepseek/deepseek-v4-pro. So Claude Code also runs on DeepSeek, and Codex on GLM.

Make a separate key for each tool and give it a daily or monthly spending limit on the keys page. If a key leaks or an agent gets stuck in a loop, only that key stops.

Which address does my tool want?

  • A base address; the tool adds /chat/completions
    https://ai.fluxcdn.cloud/api/v1
    Most tools: Cursor, Cline, Continue, Zed, JetBrains, Aider, OpenCode, Open WebUI, n8n, Dify, every OpenAI SDK, and Codex
  • An Anthropic address; the tool adds /v1/messages
    https://ai.fluxcdn.cloud/api
    Claude Code, the Anthropic SDKs, LibreChat with provider: anthropic
  • The full address of one request
    https://ai.fluxcdn.cloud/api/v1/chat/completions
    VS Code (Custom Endpoint), TypingMind, Emacs gptel, Make, Zapier, Google Sheets
  • An address without /v1; the tool adds /v1/chat/completions
    https://ai.fluxcdn.cloud/api
    Neovim CodeCompanion

If you are not sure, try the first row. A 404 that doesn't mention the model means /v1 is in the wrong place.

Coding agents

Agents that read code, run commands and change files. Pick a model that supports tools.
Claude CodeAnthropic's coding agent in the terminal, VS Code and JetBrains, on any FluxCDN model: Claude, DeepSeek, GLM, Kimi and the rest.Anthropic's software, not open source · macOS, Linux, WindowsFull guide →CodexOpenAI's coding agent in the terminal and the editor, on any FluxCDN model: GPT, Claude, DeepSeek, GLM and the rest.Apache-2.0, open source · macOS, Linux, WindowsFull guide →OpenCodeAn open-source coding agent for the terminal, set up with one opencode.json.MIT, open source · macOS, Linux, Windows (it recommends WSL)Full guide →AiderPair programming in the terminal with git; works with almost any chat model.Apache-2.0, open source · macOS, Linux, Windows (Python 3.10 to 3.12)Full guide →ClineAn open-source agent for VS Code, JetBrains and the terminal.Apache-2.0 · VS Code, JetBrains, CLIFull guide →Kilo CodeAn open-source agent that fetches the model list from FluxCDN itself.MIT · VS Code, JetBrains, CLIFull guide →Zoo CodeThe continuation of Roo Code, for VS Code.Apache-2.0 · VS CodeFull guide →ContinueOpen-source chat and agent with one config.yaml.Apache-2.0 · VS Code, JetBrains, CLIFull guide →
Qwen CodeA terminal coding agent, forked from Gemini CLI, that takes an OpenAI-compatible address.Apache-2.0 · macOS, Linux, Windows (Node.js 22+)How to connect

Put this in ~/.qwen/settings.json and set the key in FLUXCDN_API_KEY (envKey is the variable's name, not the key). Or type /auth in Qwen Code and choose Custom Provider. Test: qwen -p "Say hello".

~/.qwen/settings.json
{
  "modelProviders": {
    "openai": [
      {
        "id": "qwen/qwen3-coder",
        "name": "Qwen3 Coder (FluxCDN)",
        "baseUrl": "https://ai.fluxcdn.cloud/api/v1",
        "envKey": "FLUXCDN_API_KEY",
        "generationConfig": { "contextWindowSize": 262144 }
      }
    ]
  },
  "security": { "auth": { "selectedType": "openai" } },
  "model": { "name": "qwen/qwen3-coder" }
}
GooseAn open-source agent with a desktop app and a CLI.Apache-2.0 · macOS, Linux, WindowsHow to connect

In the app: Settings → Models → Configure providers → Add Custom Provider, type OpenAI Compatible, the address https://ai.fluxcdn.cloud/api/v1, the key and the model ids (comma-separated). In the terminal: goose configure and Custom Providers, or the file below. Goose ignores a key written in config.yaml; use an environment variable or the keychain.

~/.config/goose/custom_providers/custom_fluxcdn.json
{
  "name": "custom_fluxcdn",
  "engine": "openai",
  "display_name": "FluxCDN AI",
  "description": "FluxCDN AI (OpenAI-compatible)",
  "api_key_env": "CUSTOM_FLUXCDN_API_KEY",
  "base_url": "https://ai.fluxcdn.cloud/api/v1/chat/completions",
  "models": [
    { "name": "anthropic/claude-sonnet-5.5", "context_limit": 1000000 },
    { "name": "qwen/qwen3-coder", "context_limit": 262144 }
  ],
  "supports_streaming": true,
  "requires_auth": true
}
CrushA terminal coding agent from Charm.FSL-1.1-MIT: source-available, not OSI open source; becomes MIT after two years · macOS, Linux, WindowsHow to connect

Put this in ~/.config/crush/crushrc and set the key in FLUXCDN_API_KEY. Test: crush run "Say hello". For the Anthropic format: --type messages --base-url "https://ai.fluxcdn.cloud/api".

~/.config/crush/crushrc
provider add fluxcdn --name "FluxCDN AI" --type completions \
  --base-url "https://ai.fluxcdn.cloud/api/v1" \
  --api-key "${FLUXCDN_API_KEY:?set FLUXCDN_API_KEY}"

model add fluxcdn/anthropic/claude-sonnet-5.5 --name "Claude Sonnet 5.5" --context-window 1000000 --default-max-tokens 32000
model add fluxcdn/qwen/qwen3-coder --name "Qwen3 Coder" --context-window 262144 --default-max-tokens 32000

model large fluxcdn/anthropic/claude-sonnet-5.5
model small fluxcdn/qwen/qwen3-coder

Editors and IDEs

Chat and agents inside the editor. Completion while you type doesn't use this route (see the end of the page).
VS CodeCopilot's chat and agent without a Copilot plan, or the open-source Cline, Kilo Code, Zoo Code and Continue extensions.VS Code and the Copilot Chat extension: MIT; the Copilot service is proprietary · macOS, Linux, WindowsFull guide →CursorCursor's chat in Ask mode on FluxCDN models. Not Agent mode, not Tab completion.Proprietary; by the available guides, custom models need a paid Cursor plan · macOS, Linux, WindowsFull guide →JetBrainsAI Assistant in IntelliJ IDEA, PyCharm, WebStorm and the rest with your own key, no JetBrains AI subscription; and the Junie CLI.Proprietary; the AI Assistant plugin is free · macOS, Linux, WindowsFull guide →ZedA fast open-source editor; its agent, inline assistant and commit messages on FluxCDN models.GPL-3.0 (some parts Apache-2.0) · macOS, Linux, WindowsFull guide →
NeovimTwo popular plugins: avante.nvim and CodeCompanion.nvim.Apache-2.0 · Wherever Neovim runsHow to connect

Put the key in FLUXCDN_API_KEY. avante: the address with /v1. CodeCompanion: the address without /v1, as it adds /v1/chat/completions itself; always set api_key, or your OPENAI_API_KEY is sent instead.

lazy.nvim spec
opts = {
  provider = "fluxcdn",
  providers = {
    fluxcdn = {
      __inherited_from = "openai",
      endpoint = "https://ai.fluxcdn.cloud/api/v1",
      api_key_name = "FLUXCDN_API_KEY",
      model = "qwen/qwen3-coder",
    },
  },
},
Emacs (gptel)Chat with models from any Emacs buffer.GPL-3.0 · Wherever Emacs runsHow to connect

gptel takes the address in two parts: :host without https://, and :endpoint with the full path. gptel doesn't fetch the model list, so name the models. Test: M-x gptel.

init.el
(setq gptel-model 'anthropic/claude-sonnet-5.5
      gptel-backend
      (gptel-make-openai "FluxCDN"
        :host "ai.fluxcdn.cloud"
        :endpoint "/api/v1/chat/completions"
        :stream t
        :key (lambda () (getenv "FLUXCDN_API_KEY"))
        :models '(anthropic/claude-sonnet-5.5
                  deepseek/deepseek-v4-pro
                  qwen/qwen3-coder)))

Chat apps

ChatGPT-like apps you install or run on your own server. They fetch the model list from FluxCDN themselves.
Open WebUIA self-hosted chat app for a team or a family; it lists every FluxCDN model by itself.Open WebUI License: BSD-3 based with a branding clause; not OSI open source · Self-hosted (Docker, pip); any browserFull guide →LibreChatAn open-source, self-hosted chat app with agents and many users; set up with one librechat.yaml.MIT, open source · Self-hosted (Docker, Node); any browserFull guide →Cherry Studio, Chatbox, JanThree free chat apps for Windows, Mac and Linux that reach every FluxCDN model with your key.Full guide →Cherry StudioA desktop chat app with dozens of helpers.AGPL-3.0 · Windows, macOS, LinuxFull guide →ChatboxA chat app for desktop, mobile and the web.GPL-3.0 (Community) · Windows, macOS, Linux, iOS, Android, webFull guide →JanAn open-source desktop chat app.Apache-2.0 · Windows, macOS, LinuxFull guide →
LobeHubA web and desktop chat app with agents and plugins.LobeHub Community License (Apache-2.0 with conditions; not OSI open source) · Web (self-hosted), desktopHow to connect

Settings → Provider, Add Custom Provider: id fluxcdn, Request Format openai, Proxy URL https://ai.fluxcdn.cloud/api/v1 and the key. Then Fetch models and Connectivity Check.

AnythingLLMChat with your own documents, on the desktop or a server.MIT · Windows, macOS, Linux, DockerHow to connect

Settings → AI Providers → LLM, provider Generic OpenAI: the address, the key, the model's context size and its id. Keep the embedder on its own local option. In Docker, the same with these variables:

.env
LLM_PROVIDER='generic-openai'
GENERIC_OPEN_AI_BASE_PATH='https://ai.fluxcdn.cloud/api/v1'
GENERIC_OPEN_AI_API_KEY='flux_live_YOUR_KEY'
GENERIC_OPEN_AI_MODEL_PREF='anthropic/claude-sonnet-5.5'
GENERIC_OPEN_AI_MODEL_TOKEN_LIMIT=1000000
MstyA desktop chat app.Proprietary (has a free tier) · Windows, macOS, LinuxHow to connect

Under Model Providers add an OpenAI-compatible provider, enter https://ai.fluxcdn.cloud/api/v1 and the key, and fetch the models. The exact option names differ between Msty versions.

TypingMindA web chat app that sends requests straight from the browser.Proprietary · WebHow to connect

Models → Add custom model with these values, then Test & Save. Once per model. TypingMind wants the full address.

Add custom model
Model ID:            anthropic/claude-sonnet-5.5
API Type:            OpenAI Chat Completions API
Endpoint URL:        https://ai.fluxcdn.cloud/api/v1/chat/completions
Authentication Type: API Key via HTTP Header
Header Key:          Authorization
Header Value:        Bearer flux_live_YOUR_KEY
Page AssistA browser extension to chat beside any page.MIT · Chrome, Edge, Brave, FirefoxHow to connect

The extension's settings, tab OpenAI Compatible API, Add Provider and Custom: the address and key. “Chat with page” needs an embedding model; keep that local (Ollama).

Automation and no-code

Workflows, bots and agents you build without writing code.
n8nWorkflows, agents and Telegram bots without code; through n8n's own OpenAI credential.Sustainable Use License (fair-code; not OSI open source) · Self-hosted (Docker, npm) or n8n CloudFull guide →DifyBuild AI apps, agents and workflows with the official OpenAI-API-compatible plugin.Dify Open Source License (Apache-2.0 based, with extra conditions) · Self-hosted (Docker) or Dify CloudFull guide →
FlowiseBuild agents and flows by drag and drop.Apache-2.0 · Self-hosted or Flowise CloudHow to connect

Create an OpenAI API credential with the key. Node OpenAI Custom Model: Model Name the id (for example anthropic/claude-sonnet-5.5) and Additional Parameters → Base Path https://ai.fluxcdn.cloud/api/v1.

LangflowBuild flows and agents visually.MIT · Python, Docker, desktopHow to connect

Settings → Model Providers → OpenAI Compatible: the address and key, then Save. The models arrive by themselves; turn them on under Language Models. Or the variables OPENAI_COMPATIBLE_BASE_URL and OPENAI_COMPATIBLE_API_KEY.

ActivepiecesOpen-source automation in the style of Zapier.MIT (core) · Cloud or self-hosted (Docker)How to connect

Platform admin: AI Providers and Other (OpenAI Compatible) with the address and key. If your edition doesn't have it, use the HTTP piece with a POST to https://ai.fluxcdn.cloud/api/v1/chat/completions.

MakeMake's OpenAI app takes no address; the HTTP module does the job.Proprietary · WebHow to connect

The HTTP → Make a request module with these values and Parse response on. The answer is in data.choices[].message.content. Don't ask for a stream; Make can't read it.

HTTP → Make a request
Method:  POST
URL:     https://ai.fluxcdn.cloud/api/v1/chat/completions
Header:  Authorization: Bearer flux_live_YOUR_KEY
Body:    {"model": "deepseek/deepseek-v4-flash", "messages": [{"role": "user", "content": "…"}]}
ZapierZapier's ChatGPT app takes no address; Webhooks does the job.Proprietary · WebHow to connect

Webhooks by Zapier → Custom Request (on Zapier's paid plans) with these values and the header Content-Type: application/json. Text you map into the JSON must not carry unescaped quotes or line breaks.

Webhooks by Zapier → Custom Request
Method:  POST
URL:     https://ai.fluxcdn.cloud/api/v1/chat/completions
Header:  Authorization: Bearer flux_live_YOUR_KEY
Body:    {"model": "deepseek/deepseek-v4-flash", "messages": [{"role": "user", "content": "…"}]}

Other places

Notes, launchers, WordPress, the smart home, spreadsheets and Telegram.
ObsidianThe Copilot and Text Generator plugins for your notes.Copilot: AGPL-3.0; Text Generator: MIT · Desktop and mobileHow to connect

Copilot: Settings → Copilot → BYOK → Add a custom provider: the address and key, then Test and Save. If Test works but chat doesn't, turn on Enable CORS for that provider. Text Generator: the OpenAI Chat provider with Base Path https://ai.fluxcdn.cloud/api/v1.

RaycastThe macOS and Windows launcher; FluxCDN models in Quick AI.Proprietary; Custom Providers need Raycast Pro · macOS, WindowsHow to connect

Settings → AI → Models & Providers, section Custom Providers, Reveal Providers Config. Write providers.yaml like this; Raycast picks up the change itself. Declare tools only for models that support tools.

~/.config/raycast/ai/providers.yaml
providers:
  - id: fluxcdn
    name: FluxCDN AI
    base_url: https://ai.fluxcdn.cloud/api/v1
    api_keys:
      fluxcdn: flux_live_YOUR_KEY
    models:
      - id: anthropic/claude-sonnet-5.5
        name: Claude Sonnet 5.5 (FluxCDN)
        context: 1000000
        abilities:
          temperature: { supported: true }
          vision: { supported: true }
          system_message: { supported: true }
          tools: { supported: true }
      - id: deepseek/deepseek-v4-flash
        name: DeepSeek V4 Flash (FluxCDN)
        context: 1048576
WordPress (AI Engine)Chatbots and content in WordPress with the free AI Engine plugin.GPL-2.0 or later (free version, 3.4.8 and later) · WordPressHow to connect

Meow Apps → AI Engine → Settings, section Environments for AI: an environment of the custom OpenAI-compatible type, Endpoint https://ai.fluxcdn.cloud/api/v1 and the key. The models arrive from FluxCDN; then choose the environment for the chatbot.

Home AssistantThe smart home's voice assistant on FluxCDN models.Home Assistant: Apache-2.0 · Wherever Home Assistant runsHow to connect

The official OpenAI integration takes no address. Install Extended OpenAI Conversation from HACS: Base Url https://ai.fluxcdn.cloud/api/v1, the key, and a model with good tool support (such as openai/gpt-5.5). Then make it the conversation agent under Voice assistants.

Google SheetsA =FLUX() spreadsheet function with Apps Script.How to connect

Extensions → Apps Script. Under Project Settings → Script Properties add FLUX_KEY with your key, paste this code and save. In a cell type =FLUX("Say hello"). Custom functions must finish within 30 seconds; ask for a quick model and a short answer. Every recalculation is a paid request.

Apps Script: Code.gs
const FLUX_BASE = 'https://ai.fluxcdn.cloud/api/v1';

/**
 * Ask FluxCDN AI. =FLUX("Summarise: " & A2)  or  =FLUX(A2, "anthropic/claude-haiku-5.5")
 * @customfunction
 */
function FLUX(prompt, model) {
  if (!prompt) return '';
  const key = PropertiesService.getScriptProperties().getProperty('FLUX_KEY');
  const res = UrlFetchApp.fetch(FLUX_BASE + '/chat/completions', {
    method: 'post',
    contentType: 'application/json',
    headers: { Authorization: 'Bearer ' + key },
    payload: JSON.stringify({
      model: model || 'deepseek/deepseek-v4-flash',
      messages: [{ role: 'user', content: String(prompt) }],
      max_tokens: 400,
    }),
    muteHttpExceptions: true,
  });
  const data = JSON.parse(res.getContentText());
  if (res.getResponseCode() !== 200) return 'Error ' + res.getResponseCode() + ': ' + (data.error ? data.error.message : '');
  return data.choices[0].message.content.trim();
}
ExcelWith the GPT for Excel (Autosheet) add-in.Proprietary; custom endpoints on Business, Enterprise or pay-as-you-go · Excel (Windows, Mac, web)How to connect

The Autosheet dashboard: Custom API endpoints → Bulk models → OpenAI-compatible endpoints (Chat Completions API), the address https://ai.fluxcdn.cloud/api/v1 and the key, then Check and Save. If Check fails with that address, try https://ai.fluxcdn.cloud/api.

TelegramYour own Telegram bot, in a few lines of Python or with n8n.python-telegram-bot: LGPL-3.0 · Your server (Python 3.11+)How to connect

pip install python-telegram-bot openai, put the bot token from BotFather in TELEGRAM_BOT_TOKEN and the key in FLUXCDN_API_KEY, and run it. A Telegram message holds at most 4096 characters. Without code: in n8n, connect a Telegram Trigger to an AI Agent.

bot.py
import os
from openai import AsyncOpenAI
from telegram.ext import Application, MessageHandler, filters

ai = AsyncOpenAI(base_url="https://ai.fluxcdn.cloud/api/v1", api_key=os.environ["FLUXCDN_API_KEY"])

async def reply(update, context):
    answer = await ai.chat.completions.create(
        model="deepseek/deepseek-v4-flash",
        messages=[{"role": "user", "content": update.message.text}],
    )
    await update.message.reply_text(answer.choices[0].message.content[:4096])

app = Application.builder().token(os.environ["TELEGRAM_BOT_TOKEN"]).build()
app.add_handler(MessageHandler(filters.TEXT & ~filters.COMMAND, reply))
app.run_polling()

In your own code

The official OpenAI or Anthropic SDK, or a framework built on them. Only the address and the key change.

ChatGPT, Windsurf and Gemini CLI

These have no place for another service's address or key. Use this instead:
ChatGPT
The ChatGPT app doesn't take another service's key or address. For a ChatGPT-like app, FluxCDN Chat has every model with nothing to set up. Or run one of the chat apps above.
Windsurf (Devin Desktop)
It has no setting for your own key or address. Install the Cline, Kilo Code, Zoo Code or Continue extension inside it and connect that (VS Code guide).
Gemini CLI
It only talks to Google's own API. Try Qwen Code, which started as a fork of Gemini CLI and takes an OpenAI-compatible address, or OpenCode, Crush or Goose.

When it doesn't work

Error 401 or invalid_api_key
The key wasn't copied whole, was revoked, or went in the wrong field. It starts with flux_live_ and has no spaces. If the tool reads it from an environment variable, open a new terminal or window after setting it.
Error 402, insufficient_credit (in Codex: insufficient_quota) although you have credit
Before each request, the most it could cost is set aside from your credit. When the answer ends, only the real cost is taken and the rest is released. A tool that asks for a very long output on every request sets aside a large amount. Lower the tool's output limit (for example to 32000 tokens) or add credit.
Error 404 or model_not_found
The model id must be exactly as on the models page, with the maker's prefix: anthropic/claude-sonnet-5.5, not claude-sonnet-5.5. Lower case and dots matter. Copy the id rather than typing it.
A 404 that doesn't mention the model
The base address is wrong. Most tools want https://ai.fluxcdn.cloud/api/v1 and add the rest themselves; don't write /chat/completions twice and don't leave out /v1. See the which address table.
An error about previous_response_id or OpenAI's built-in tools (web search, file search, code interpreter)
Some tools switch to the Responses API without saying so. FluxCDN has it, but without memory (the whole conversation is sent each time) and without the tools that run on OpenAI's servers. Set the tool to Chat Completions: in n8n turn off Use Responses API, in Continue useResponsesApi: false, in the Vercel AI SDK openai.chat(...), in LangChain use_responses_api=False.
Error 429: rate_limit_exceeded or budget_exceeded
Either requests per minute went over the limit (wait a little; most tools retry by themselves), or this key's spending today or this month reached its limit. See it on API keys under “Limits”.
A CORS error or “Network error” in browser apps
FluxCDN answers requests made straight from a browser, so CORS is no longer in the way. If you still see this error, check the address and the key; a browser sometimes shows a 401 or 404 as a network error. Never put your key in a page other people open.
The agent calls no tools, or the model is missing from the agent's list
Pick a model that supports tools; it has the Tools label on the models page. Some tools need that switched on for the model (the steps above). Routers are not recommended for agents.
Error 502, 503, 504 or 529
The model's service was briefly unavailable. Send again; most tools do it themselves. A request that fails before answering costs nothing. If it lasts, switch to another model for a while.

What doesn't work this way

  • No embeddings. Knowledge bases, file search and code indexing in these tools need a separate embedding model. Most of them have a local one that is enough.
  • Pictures and sound are made through Chat Completions, not /images or /audio. The apps' “generate image” and “text to speech” buttons usually call those paths and don't work.
  • Completion while you type (Cursor's Tab, Copilot's grey suggestions, Zed's edit predictions) doesn't use FluxCDN; it needs a /completions path we don't have.
  • Tools that run on the model maker's servers don't work: WebSearch in Claude Code, web_search in Codex, OpenAI's file search and code interpreter. Reading a given page (WebFetch in Claude Code) works.
  • The Responses API keeps no state: previous_response_id isn't accepted and the whole conversation is sent each time (as Codex does).