Choose Your AI Provider

Phyllax supports 8 AI providers. You need at least one to power monitoring and the AI assistant. Your provider key goes from your machine directly to the provider — Phyllax never sees it, because there is no Phyllax server.

Free Options

Google Gemini (Recommended for new users)

1. Go to Google AI Studio
2. Click Create API Key
3. Copy the key
4. In Phyllax Settings → select Gemini → paste the key → Save

TIP

Gemini's free tier is generous — enough for most personal use. No credit card required.

Ollama (100% local & private)

1. Download Ollama from ollama.com
2. Install and run it
3. Phyllax detects Ollama automatically — no API key needed
4. The setup wizard will download a model (default: qwen3:32b — best open-source model for tool calling)

TIP

Ollama runs entirely on your machine. No data leaves your computer. Free forever. Smaller machines can swap to qwen2.5:7b (~4GB) or qwen2.5:14b from Settings.

Paid Options

Anthropic Claude

1. Go to console.anthropic.com → API Keys
2. Create a new key
3. Paste in Phyllax Settings → Anthropic

OpenAI

1. Go to platform.openai.com/api-keys
2. Create a new key
3. Paste in Phyllax Settings → OpenAI

xAI Grok

1. Go to console.x.ai
2. Create a new key
3. Paste in Phyllax Settings → Grok

Azure OpenAI

Requires an Azure subscription with an OpenAI resource deployed. Enter your endpoint, API key, deployment name, and API version in Settings.

AWS Bedrock

Requires an AWS account with Bedrock model access enabled. Enter your access key, secret key, region, and model ID in Settings. Phyllax uses the Converse API with AWS Signature V4 (Node crypto only — no AWS SDK).

Bring Your Own — OpenAI Compatible Endpoint

The OpenAI Compatible provider lets you point Phyllax at any OpenAI-protocol endpoint. This unlocks self-hosted, enterprise, and alternative model providers without a custom integration:

Use case	Example endpoint
Together AI	https://api.together.ai/v1
Groq	https://api.groq.com/openai/v1
Fireworks	https://api.fireworks.ai/inference/v1
Mistral La Plateforme	https://api.mistral.ai/v1
Self-hosted vLLM	http://your-server:8000/v1
Self-hosted LM Studio (server mode)	http://your-server:1234/v1
Your company's internal inference gateway	https://ai.your-company.internal/v1

To configure:

1. In Phyllax Settings → select OpenAI Compatible
2. Enter the Base URL (the /v1 root that handles /chat/completions)
3. Enter an API key (if the endpoint requires one — leave blank for unauthenticated local servers)
4. Enter the Model ID the endpoint expects (e.g. Qwen/Qwen3.5-397B-A17B, mistral-large-latest, meta-llama/Llama-3.3-70B-Instruct-Turbo)
5. Click Test connection → Save

Why this matters for privacy-conscious teams

Pointing OpenAI Compatible at an internal inference gateway means your prompts, integration data, and tool calls all stay inside your network. Phyllax becomes a thin local client that orchestrates MCP and routes inference to your infrastructure — no third-party API in the loop.

Using Different Providers for Different Tasks

In Settings → Advanced, you can set separate providers for:

• Assistant — powers the AI chat, suggested replies, and reply drafting
• Scheduler — powers the background monitoring checks

For example, you could use Gemini (free) for the scheduler that runs every few minutes, and Claude (higher quality) for the assistant chat when you need more nuanced responses. This helps manage costs — the scheduler makes many small calls, so using a free or local provider there saves money.

To set this up:

1. Configure at least 2 providers in Settings
2. Click Use this on your main provider
3. Go to Advanced section → set assistant and scheduler preferences

Common pattern: local scheduler + cloud assistant

Run Ollama or vLLM locally for the scheduler (every-2-minute checks against your inbox, GitHub, Stripe, etc. stay completely offline), and keep Claude or GPT for the chat assistant where you want top-tier reasoning. Best of both worlds.