OpenRouter
Default breadth and curated model choices. One key, most of the models worth having.
AI Sources
Any model. Your keys. No lock-in.
Attyn is the layer, not the intelligence. A supported cloud provider or a local model, set per workspace, with the keys held by macOS rather than by us.
Supported sources
Six sources. One place to choose.
A local Ollama model for private work, a supported cloud model for hard problems, switch anytime.
OpenAI
OpenAI-compatible streaming and tool calls. Your own key, billed by them, not us.
Groq
Fast OpenAI-compatible streaming and tool calls. The one to pick when latency matters.
Anthropic
Text-first reasoning through the native Messages API, not a compatibility shim.
Gemini
Text-first reasoning through native Gemini streaming, on your own Google key.
Ollama
Local models from the running Ollama server. Nothing leaves the machine on this one.
Keys and workspaces
Keys held by your Mac, not by us.
Chosen per workspace, sealed in the Keychain.
Point each workspace at the source you trust and switch anytime. Provider keys are sealed in the macOS Keychain, not stored by us.
AI-provider usage and charges are managed separately through the provider you select.
Cloud requests follow the selected provider’s policy.
Your keys, your models, nothing locked in.
Get Attyn
The layer is ours. The intelligence is yours.
Attyn is available for macOS today, and the Windows waitlist is open.
macOS today, Apple silicon. M1 and every M-series chip after it. Windows waitlist open.