Skip to content

Wire your tools to Fluxo

Every client talks to the API host https://api.fluxorelay.com. The website domain does not answer API calls.

VS Code

Claude Code for VS Code runs the same CLI underneath, so it picks up the two variables below from the environment VS Code was started in.

export ANTHROPIC_BASE_URL=https://api.fluxorelay.com
export ANTHROPIC_AUTH_TOKEN=flx_your_key_here
claude

The Codex extension reads ~/.codex/config.toml, the very file the Codex CLI uses; the terminal section below has a ready copy.

Extensions with an "OpenAI compatible" provider (Cline, Roo Code, Continue) need three fields: base URL https://api.fluxorelay.com/v1, your Fluxo key, and a model id from the pricing page.

Base URL : https://api.fluxorelay.com/v1
API key  : flx_your_key_here
Model    : gpt-6.1-sol

JetBrains IDEs

The Claude Code plugin drives the CLI inside the IDE terminal: put the two exports in your shell profile and restart the IDE so it inherits them.

Assistant plugins that accept a custom OpenAI endpoint take https://api.fluxorelay.com/v1 as the URL and your Fluxo key as the token.

Cursor

In Cursor's model settings, paste your Fluxo key as the OpenAI API key, switch on the base URL override and set it to https://api.fluxorelay.com/v1, then add the model ids you want, such as gpt-6.1-sol.

Claude models are not reachable through Cursor's OpenAI override; use Claude Code for them (note in the terminal section).

Terminal

Command-line agents read their settings from the environment or a dotfile. One file per tool below.

Claude Code and Claude Desktop

Set two environment variables, then launch claude. The base URL carries no /v1 suffix.

export ANTHROPIC_BASE_URL=https://api.fluxorelay.com
export ANTHROPIC_AUTH_TOKEN=flx_your_key_here
claude

PowerShell variant:

$env:ANTHROPIC_BASE_URL = "https://api.fluxorelay.com"
$env:ANTHROPIC_AUTH_TOKEN = "flx_your_key_here"
claude

Claude models are served only to the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. That gate sits at the upstream provider. GPT, Gemini and every other family work from any OpenAI-compatible client or SDK.

Codex CLI

Codex reads ~/.codex/config.toml. The env_key line names the variable that holds your Fluxo key.

.codex/config.toml

# ~/.codex/config.toml  (Codex CLI and the Codex IDE extension both read it)
model = "gpt-5.6-sol"        # or any other GPT id from the Fluxo price list
model_provider = "fluxo"

[model_providers.fluxo]
name = "Fluxo relay"
base_url = "https://api.fluxorelay.com/v1"
env_key = "FLUXO_API_KEY"   # export FLUXO_API_KEY=flx_... before launching codex
wire_api = "responses"       # Fluxo passes /v1/responses straight through
export FLUXO_API_KEY=flx_your_key_here
codex   # reads ~/.codex/config.toml

Gemini CLI and the Gemini API

Point GOOGLE_GEMINI_BASE_URL at the bare API host (do not append /v1beta) and put your key in GEMINI_API_KEY, then run gemini.

export GOOGLE_GEMINI_BASE_URL=https://api.fluxorelay.com
export GEMINI_API_KEY=flx_your_key_here
gemini
curl "https://api.fluxorelay.com/v1beta/models/gemini-3.8-flash:generateContent" \
  -H "x-goog-api-key: flx_your_key_here" \
  -H 'Content-Type: application/json' \
  --data-raw '{"contents":[{"role":"user","parts":[{"text":"ping from my editor"}]}]}'

Gemini speaks only its native protocol on Fluxo. Stream with :streamGenerateContent?alt=sse; /v1/chat/completions will refuse Gemini models.

Gemini model ids:

  • gemini-3.1-pro
  • gemini-3.8-flash
  • gemini-4-argon

cURL straight from the shell

Use https://api.fluxorelay.com/v1 as base_url and your Fluxo key as api_key. Chat calls go to /v1/chat/completions.

curl https://api.fluxorelay.com/v1/chat/completions \
  -H "Authorization: Bearer flx_your_key_here" \
  -H 'Content-Type: application/json' \
  --data-raw '{"model":"gpt-6.1-sol","messages":[{"role":"user","content":"ping from my editor"}]}'

GPT model ids worth starting with:

  • gpt-6.1-sol
  • gpt-6-astra
  • gpt-5.6-sol

Keys, errors and billing

OpenAI-style calls carry the key as a Bearer token. Anthropic-style calls carry it in x-api-key. Gemini calls use x-goog-api-key.

Each call is billed on the token counts the provider reports. HTTP 402 means your credit is at zero or your pass allowance is spent: refuel, then retry.