Connect your AI provider

Provider comparison

Anthropic (Claude)

Claude is the default and recommended provider. It produces detailed reasoning and conservative trade assessments.

OAuth (use your existing subscription)

If you already have a Claude subscription (Pro, Team, or Enterprise), you can authenticate via OAuth to use your existing subscription. The setup wizard guides you through the sign-in flow.

API key (bring your own)

1. 1.Go to console.anthropic.com and create an API key.
2. 2.In the setup wizard, select "Anthropic" as your provider and "API Key" as auth type.
3. 3.Paste your key. It is encrypted locally and never sent to BrokerBridge servers.

ai_provider:
  name: anthropic
  auth_type: api_key
  model: claude-sonnet-4-6

OpenAI

OpenAI provides GPT-5.4 and other models. You need your own OpenAI account -- either use your existing subscription via OAuth, or create an API key at platform.openai.com.

1. 1.Go to platform.openai.com/api-keys and create a key.
2. 2.Select "OpenAI" in the setup wizard and paste your key.

ai_provider:
  name: openai
  auth_type: api_key
  model: gpt-5.4

OpenRouter

OpenRouter provides access to many models through a single API. Useful if you want to switch between different models without changing provider configuration.

1. 1.Go to openrouter.ai/keys and create an API key.
2. 2.Choose a model from the OpenRouter model list. Set it in your config.

ai_provider:
  name: openrouter
  auth_type: api_key
  model: anthropic/claude-sonnet-4-6

Google AI

Google AI provides Gemini models through AI Studio. Fast response times and competitive pricing.

1. 1.Go to aistudio.google.com/app/apikey and create an API key.
2. 2.Select "Google" in the setup wizard and paste your key.

ai_provider:
  name: google
  auth_type: api_key
  model: gemini-2.0-flash

Local models (Ollama)

Run an AI model entirely on your own machine with Ollama. Nothing leaves your computer, there is no per-token cost, and it works offline. The trade-off is speed and quality: a local model depends on your hardware, and smaller models are less capable than the frontier cloud models above. Recommended for privacy, cost control, or air-gapped setups — not for the fastest possible responses.

1. 1.Install Ollama from ollama.com/download and make sure it is running (the menu-bar icon, or ollama serve).
2. 2.Pull a model in a terminal, for example ollama pull ornith:35b. This downloads once (~21 GB) and is cached locally.
3. 3.In BrokerBridge, open the model picker and choose Ollama (local), then pick the model you pulled. No API key is required.

ai_provider:
  name: ollama
  auth_type: api_key   # a placeholder key; the local server ignores it
  model: ornith:35b
  base_url: http://localhost:11434/v1

The dropdown shows a measured latency estimate next to each model once you have used it, so you can see how your local model compares to the cloud providers on your own hardware.

How AI decisions work

When BrokerBridge identifies a trade setup, it sends the AI a structured prompt containing:

• •Price action data (recent OHLCV bars and extracted features)
• •Evidence snapshots from data source plugins (scanner signals, flow data)
• •Current portfolio state (open positions, daily P&L, exposure)
• •Risk parameters (your configured limits)

The AI returns a structured decision: approve or reject, with a confidence score (0-1), reasoning summary, suggested position size, and risk/reward ratio. This decision is recorded in the decisions CSV and presented alongside the original proposal.