Models

The Models page lets you manage the AI models your agents use. Synthcore is provider-agnostic — bring your own API keys and choose the models that work best for your team.

Supported providers

Provider	Models	Use case
Anthropic	Claude Opus, Sonnet, Haiku	General development, code review, planning
OpenAI	GPT-4o, GPT-4o-mini	Alternative coding, broad task support
Google	Gemini 2.5 Pro, Flash	Fast iteration, large context tasks
MiniMax	MiniMax models	Flat-rate coding (used by engine runtime)
xAI	Grok models	Alternative coding and reasoning

Providers are added as they become available. The dispatcher supports automatic fallback across providers — if your primary model is unavailable, agents switch to the next provider in the chain without interruption.

Adding API keys

1. Go to Models in the sidebar
2. Click Add Provider
3. Enter your API key for the provider
4. The key is encrypted and stored securely — never shared between projects or customers

You can add keys for multiple providers. Agents will use whichever provider is configured for their role.

Model configuration

Each project can be configured with:

• Primary model — The default model agents use for most tasks
• Fallback chain — Backup models used automatically if the primary is unavailable or rate-limited (e.g., Claude Sonnet → GPT-4o-mini → Gemini Flash)
• Per-agent overrides — Assign specific models to agents based on their role. A code-review agent might benefit from a larger context window, while a test-writing agent can use a faster, cheaper model

Viewing usage

The models page shows:

• Which models are active across your projects
• Current usage levels and token consumption per model
• Rate limiting or availability issues from providers
• Cost per model — see the Costs page for detailed breakdowns

Best practices

• Start with defaults — Synthcore ships with sensible model assignments. Customize once you see usage patterns
• Use fallbacks — Provider outages happen. A fallback chain keeps agents productive
• Match model to task — High-capability models for architecture and review, fast models for routine tasks