AI agents

AI steps vs. agents

• AI step — a deterministic step in your flow. You give it a prompt and inputs, and it returns text or structured data you map into later steps. Best when you know exactly where AI fits.
• Agent — you give it a goal and a set of tools, and it reasons through the steps on its own, calling tools until the goal is met. Best for open-ended tasks like triage, research, or support.

Choosing a provider and model

Configure one or more AI providers for your workspace, then pick a model per step or agent. AgentFlow10 supports leading providers, and we recommend the latest and most capable models for reasoning-heavy tasks.

Bringing your own provider key means AI usage is billed by your provider at their rates, and your prompts and data go directly to the provider you choose.

Structured output

Ask an AI step to return data in a fixed shape so downstream steps can rely on it. You describe the fields you want, and the model returns JSON matching that structure — ideal for classification, extraction, and routing.

{
  "category": "billing",
  "priority": "high",
  "summary": "Customer was double-charged for the Pro plan."
}

Giving agents tools

An agent becomes useful when you give it tools. Tools can be:

• Actions from any connected app (send an email, create a ticket).
• Your own flows, exposed as callable tools.
• Tables and data lookups within your workspace.

The agent decides which tools to call and in what order, using the data it gathers along the way.

Guardrails & review

• Scope each agent to only the tools it needs, so it cannot take actions outside its job.
• Add a human approval step before high-impact actions such as sending external messages or moving money.
• Every agent run is logged step by step, so you can audit exactly what it did and why.

MCP support

AgentFlow10 supports the Model Context Protocol, letting external AI clients securely use your flows and tools as capabilities — and letting your agents consume MCP tools in return.