Beta Limitations

Canoryn is in active beta. The goal is to make local-first Mac automation useful and trustworthy before expanding the product surface.

What Works Best Today

Canoryn is best suited for:

• building workflows with Chat, Build with chat, or by wiring nodes by hand,
• Browser Nodes and research boards for live multi-page reading and comparison,
• voice / chat / schedule / app-event triggers for personal Mac automation,
• running simple app and system actions with explicit permissions,
• using the CLI & MCP so other AI tools can author or run enabled workflows,
• sharing workflow ideas through .cryn files.

What May Still Change

During beta, these areas may change between releases:

• node names and action schemas,
• workflow file format details,
• AI provider settings,
• memory behavior,
• app permissions and onboarding flow,
• store/discovery publishing flow,
• Build with chat phrasing and clarification UX,
• which coding agents appear as dispatch targets.

Known Practical Limits

• Some actions require macOS permissions before they work.
• Cloud AI providers require your own API key (stored in the macOS Keychain).
• Local AI requires Ollama or another supported local model setup.
• Complex UI automation can fail if the target app changes its interface.
• Not every action listed in older examples may be available in every beta build.
• Builds are still beta and may not be notarized — see Installation for first-open steps.
• Multi-agent supervision (Mission Control) remains experimental.

Trust and Safety Expectations

• Keep important workflows simple and test them before relying on them.
• Review actions that can modify files or control apps.
• Avoid giving broad file permissions unless your workflow needs them.
• If a workflow uses a cloud AI provider, assume prompt/context data is sent to that provider.

How to Help

If something breaks or feels unclear:

• use Support, or
• open a docs correction using Edit this page on GitHub.