Overview

Agentastic separates two ideas: providers (where models come from) and models (the specific brains you run). This page explains how they fit together; Configuring Providers covers the setup steps.

Providers

Agentastic has first-class support for three providers:

• Agentastic — the built-in cloud provider, included with your plan. No API key required, and it gives you access to a large catalog of frontier models from many vendors through one connection.
• Anthropic — connect directly with an API key or your Claude subscription to run Claude models.
• OpenAI — connect directly with an API key or your ChatGPT subscription to run GPT models.

You can enable any combination. The models a provider offers only appear once that provider is connected and enabled.

Models

A model is identified as provider:model, for example anthropic:claude-opus, openai:gpt-5, or agentastic:auto. In the launcher you select a model for a single request with the /model: tag:

/model:opus  Summarize this contract and flag risky clauses

See Choosing Models for the full catalog and guidance on which to use when.

Per-role defaults

Rather than picking a model every time, Agentastic lets you set a default model for each role:

Set these in Settings → Models and Providers. Agentastic picks the right default automatically based on what you're doing.

Automatic routing

The agentastic:auto model lets Agentastic choose the best model for each request based on the task. It's the simplest way to always be on a capable model without thinking about it.

Bundles (compare models)

A bundle runs your prompt against several top models at once so you can compare:

• agentastic:frontier — top general-purpose models
• agentastic:imagine-frontier — top image-generation models
• agentastic:eastern-frontier — top models from Eastern labs

Trigger a comparison with Bundle mode (⌘2) or /model:frontier. See Launcher Modes.

Next

• Configuring Providers — connect credentials and test
• Choosing Models — the full model catalog