Integrations

A flow can reach past its own canvas in two ways: it can use the web, and it can use the servers your workspace connects.

Web use

Two nodes in the palette read the open web — one fetches a page and returns its text, the other searches and returns results to follow. Drop either onto the canvas and it works; there's nothing to configure.

In an agent, the pair is one switch: turn on Web use under Capabilities and the agent decides at run time when to search and what to open.

Servers your workspace connects

Anything else your team runs comes in as an MCP server. Connect one in your workspace's MCP servers settings, and its tools become available to the agents in that workspace. Open an agent's Capabilities panel and switch on the servers that agent may use; each row shows how many tools the server offers.

zv1 can use the same servers. Each server carries its own setting for that:

• Off — zv1 can't use it.
• Ask — every call waits for your approval.
• Auto-reads — the tools the server marks read-only run without asking; everything else waits for your approval.

Agents that use tools

An agent doesn't have to be wired to a fixed destination. Alongside web use and connected servers, you can attach other flows as tools, and the agent chooses at run time which to call. That's the difference between a flow that always posts to one place and an agent that works out what to do. See Nodes for the building blocks.