AI agents
Configure an AI Agent node: its model, its tools (HTTP, MCP servers and code), and the reasoning loop.
The AI agent node is an LLM that reasons over its input and calls tools in a loop until it has an answer. You give it a model and a set of tools; it decides which tools to call and when.
Other AI nodes sit alongside it: LLM (a single prompt to a model, no tools), Lucid and Hand to coding agent; see Nodes. This page covers the full agent.
How it's wired
An AI agent is one step. Its model and tools are part of its settings, not steps of their own. It needs exactly one model, and any number of tools.
- ModelThe provider, model, and sampling settings the agent reasons with.
- ToolsHTTP, MCP, and code tools the agent can call.
- LoopThe agent reasons, calls tools, and repeats up to a set number of iterations.
Model
Pick one model. It sets:
- Provider and model: any model you've added on the Models page (OpenAI, Anthropic, Gemini, and the rest).
- Sampling: temperature, max tokens, top-p, and top-k.
You add and manage the underlying credentials once under Models; here you just pick which one this agent uses.
Tools
Add any number of tools in the agent's settings. The agent is told what each tool does and calls them as needed:
| Tool | What the agent can do |
|---|---|
| HTTP request | Call an API, with no auth, a bearer token, basic auth or a header. |
| MCP server | Use tools from a remote MCP server, added right here. Pick the subset of its tools to expose. |
| Code | Run Python the agent calls with arguments: args in, result out. |
Group related tools together so the agent has a clear toolset.
The loop
When the agent runs, it reasons with the model, optionally calls one or more tools, reads the results, and continues until it produces a final answer or hits its max turns (8 by default). Turn on Return tool calls in the output to pass each call on too. The whole thing is one node in your workflow; its output flows to the next node like any other.
When to use which AI node
- AI agent: the task needs the model to decide and act across several steps or tools.
- LLM: you just need one prompt-to-text transform (summarize, classify, rewrite).
- Lucid: the question is about your own traces, evals, alerts or issues.
Next
- Models to add a provider
- Nodes for the rest of the node types
- Integrations for app actions