15 providersBYOK · keys in the OS keychain

Bring the models you already use.

Fifteen providers, side by side. Your API keys are stored only in the OS keychain — never written to disk, never sent to a Furx server. Every call goes straight from your machine to the provider. Furx is never in the request path.

Five ways in.

— mix freely

One key, many models

fastest start

A $10 OpenRouter deposit unlocks 300+ models — the fastest way to run Council Mode with six distinct model families.

OpenRouter

Free tiers

$0

Separate free accounts, one key each. Daily quotas vary per provider; Furx orchestrates whatever you connect.

CerebrasGroqMistralSambaNovaGoogle Gemini AI Studio

Paid APIs, direct

your spend

Your billing, your rate limits, your usage caps. Calls go straight from your machine to the provider — Furx never proxies and never marks up.

AnthropicOpenAIGemini API

Local inference

$0 · private

Auto-detected on their default ports. With a local-only council, model inference never leaves your machine.

OllamaLM Studiollama.cppvLLM

Gateways & custom

org governance

Point Furx at your own gateway. Spend tracking, allow-listing, audit — any endpoint that speaks OpenAI-compatible JSON works.

LiteLLMCustom OpenAI-compatible

Add or rotate keys in Settings → Providers, or run the Furx Connect wizard on first launch. Free-tier quotas change over time — check each provider's current terms.

No proxy. Ever.

Keychain, not disk

Keys live in the OS keychain — macOS Keychain, Windows Credential Manager, libsecret on Linux. Never plaintext on disk, never in telemetry.

Direct calls

Every request goes from your machine to the provider you chose. There is no Furx-operated backend in the middle, so there is nothing to mark up.

Verify it

Every dispatch lands in a local append-only SQLite log — an audit trail the agent can't rewrite. The core is Apache-2.0 — read the source. Security page →

Connect one provider. Council them all.