Workflows
70+ one-tap AI tasks. Run offline. Trigger via Siri, widgets, share sheet, or a URL.
Avyra for iPhone and iPad
Tap. Your AI runs.
70+ on-device AI workflows for iPhone and iPad. No accounts, no servers, no analytics.
iOS 26.4+iPhone & iPadApple Intelligence ready
What's inside
AI you run, not just chat with.
Four building blocks. One universal app. Same on iPhone and iPad.
Skills
Author Anthropic-style skill bundles. Attach them per-thread in chat or per-step in a workflow.
Plugins
Sandboxed JavaScript plugins. 13 ship by default. Author your own in the in-app Runestone editor.
On-device
Apple Intelligence by default. MLX-backed models on demand. Optional OpenAI, Anthropic, or Gemini with your own keys.
Open source
Powered by Aria — an open-source agent SDK
Avyra is built on Aria, a composable on-device agent runtime for Apple platforms. The same SDK that powers Avyra's workflows, skills, and JS plugin runtime is free for any developer to build with.
Native Swift, type-safe end-to-end
No JavaScriptCore bridge for the core.
Provider-agnostic
Runs on FoundationModels, MLX, OpenAI, Anthropic, Gemini — the same `LLMProvider` protocol.
Composable runtime
Workflow runtime, capability broker, skill bundles, JS plugin runtime — all reusable from your own app.
Workflow gallery
Eight packs. Seventy plus tiles.
A sample of what ships with the app.
Daily Essentials
Morning Brief
Reads your calendar and reminders, writes the brief, hands you the day in a glance.
Productivity
Email Drafter
Bullet points in, polished message out. Pick a tone, set the audience, send.
Developer
Code Reviewer
Paste a diff, get review notes. Convention-aware. Runs on-device when you want it to.
Privacy
Your data stays on your device.
Avyra ships with no accounts, no telemetry, and no server-side state. Every line of source for the runtime is published.
No tracking, no analytics
Zero third-party SDKs. Zero outbound telemetry. Open the network inspector on a fresh install — nothing leaves the device unless you wire a step that needs to.
Keys live in the Keychain
Provider tokens (OpenAI, Anthropic, Gemini, MCP) are written to the iOS Keychain. Optionally biometric-gated. Never proxied through us.
Open Aria SDK
The agent runtime, workflow engine, and skill system are all part of the open-source Aria SDK. Audit it. Fork it. Replace it.
Apple Intelligence by default
Workflow LLM steps use Apple's on-device Foundation Models by default. The model never sees your prompt over the wire.
FAQ
A few common questions.
Full answers, including the data and HealthKit specifics, live on the FAQ page.
What is Avyra?
An on-device AI assistant for iPhone and iPad. It runs language models, voice recognition, and text-to-speech locally. Chat by text, speak via voice mode, attach images, extend with tools and plugins, and build multi-step workflows you can trigger from Shortcuts, Siri, or a URL.
Does Avyra send my data anywhere?
By default, no. The only outbound traffic is what you trigger yourself — downloading a model, enabling the HTTP tool, installing a plugin that makes web requests, configuring a server LLM provider, or wiring an MCP step.
Which models can I use?
Apple Foundation Models (the on-device default), MLX models you can download in-app (Qwen, Llama 3.2, Gemma), and server providers — OpenAI, Anthropic, and Gemini — with your own API key.