Episode 7 — Integrations & the outside world
The doors out of the workspace — connectors whose operations and allowed hosts you can read, the deep-research payoff of a bound integration, MCP servers with per-tool approval flags, and fail-closed egress.
1 min read
Your workspace does not live alone. This episode walks the doors to the outside world and the discipline built into each one: a connector you can read before you open it, the capability that lights up when an integration is bound, MCP tools that arrive with their own approval flags, and a sandbox network that answers no by default.
What the episode shows
| At | Scene |
|---|---|
| 0:13 | The catalog: connect once, the whole workspace borrows it |
| 0:30 | Reading the door: operations and allowed hosts, before anything runs |
| 0:46 | The payoff: deep research exists because Tavily is bound |
| 1:00 | MCP: your own tools, served to agents like native ones |
| 1:15 | Per-tool approval flags — native-looking is not native-trusted |
| 1:31 | The last door: sandboxed code, default-deny egress, fail-closed |
| 1:51 | The pattern at every door |
Where to go next
The integrations overview covers connecting and sharing connectors; MCP servers the protocol and its approval flags. For the network boundary, read the run-code policy — and for what a bound integration unlocks, revisit deep research.