Overview
How might we design context portability so people can trust and act on AI output?
When to use
- Essential for cross-app agents, tool-calling assistants, and multi-step workflows where users benefit from inspecting and editing the context payload that travels across each boundary.
When to skip
- Single-app tools with no ecosystem to port into.
- Highly sensitive contexts that must not leave a compliance boundary.
- Throwaway chats where portability adds ceremony without reuse.
Rules
Opaque blobs users cannot inspect before export.
Portable packages that silently include secrets.
One-way lock-in with import but no export.
Losing citations and permissions metadata in transit.
Evidence
| Product | Implementation |
|---|---|
| Model Context Protocol | Standardized context and tool interfaces across clients. |
| ChatGPT | Connectors and shared context across GPT surfaces. |
| Zapier | Mapped fields that carry data between automations. |
| Apple Shortcuts | Variables passed across multi-app shortcut steps. |