🇬🇧 EN

⚙️ Model Config

A configuration node — not executed — that specifies which AI model and provider an Agent node uses in graph mode. Connect its output edge to the model_in port of an Agent node. Every Agent node in graph mode requires exactly one Model Config connection.

Category: AI & Agents · Type identifier: model_config · Graph mode only

Overview

In graph mode, the way an Agent node gets its model settings is different from linear mode. Instead of configuring the provider, model, temperature, and token limit directly inside the Agent node's drawer, you create a separate Model Config node and draw an edge from it to the Agent.

This separation has several benefits:

Clarity — the graph canvas shows at a glance which model each agent uses, without needing to open every node's drawer.
Reusability — you can connect the same Model Config to multiple Agent nodes, ensuring they all use the same settings without duplicating configuration.
Swappability — changing the model for all connected agents is a single edit in one place.

The Model Config node carries no state and produces no output at runtime. It is purely a configuration carrier. When the workflow executes, Flusso reads the model settings from the connected Model Config node and applies them to the Agent before it runs.

Graph mode requirement: Every Agent node in graph mode must have exactly one Model Config node connected to its model_in port. If this connection is missing, the workflow will fail pre-run validation before any steps execute.

Configuration

Field	Status	Description
Provider	Required	Select one of your configured AI Providers. You can add providers in Settings → AI Providers. The provider must support the model you enter below.
Model	Required	The model identifier string, for example `gpt-4o`, `claude-sonnet-4-6`, `llama3.2`, or `mistral-large`. This must exactly match the model name your provider accepts.
Temperature	Optional	Controls response randomness on a scale from 0 to 2. Low values (0–0.3) produce consistent, focused output. Higher values (0.7–1.5) produce more varied, creative responses. Default is 0.7.
Max Tokens	Optional	The maximum number of tokens the model may generate in its response. Leave blank to use the model's built-in default. Setting a value helps control response length and manage API costs.

Output Data

No runtime output. The Model Config node does not execute and does not produce any output fields. It carries configuration data only. You cannot reference {{ model_config.output.* }} in other steps.

Example Usage

Connecting a Model Config to an Agent in graph mode

Open the graph editor and add a Model Config node. Find it in the node sidebar under AI & Agents. Drag it onto the canvas near the Agent node it will serve.
Configure the node. Open the Model Config drawer by clicking the node. Select your Provider (e.g. OpenAI), enter the Model (e.g. gpt-4o), and set Temperature to 0.2 for a factual, consistent assistant.
Draw the edge. Click the output port on the Model Config node and drag to the model_in (M) port on the Agent node. The port is marked with a small M indicator on the agent card.
Save the workflow. The Agent node will now use the connected Model Config's settings at runtime. If you change the model in the Model Config node, all connected agents are updated.

Sharing one Model Config across multiple agents

If your workflow has several Agent nodes that should all use the same model, you only need one Model Config node. Draw edges from that single Model Config to each Agent's model_in port. Changing the temperature or model in one place updates all of them simultaneously.

Tips & Notes

One Model Config per Agent. Each Agent node accepts exactly one model_in connection. You cannot connect two different Model Config nodes to the same agent. If you want two agents to use different models, create two separate Model Config nodes.
Model identifiers are case-sensitive. Enter the model name exactly as your provider expects it. Refer to your provider's documentation for the correct identifier strings.
This node is linear-mode equivalent. In linear mode, the provider, model, temperature, and max tokens fields appear directly in the Agent node's drawer. Model Config is the graph-mode equivalent of those fields — same settings, different form factor.

Related Nodes

Agent — the node this Model Config configures.
Memory — connects a memory backend to an Agent node in graph mode.
MCP Tool — connects external tools to an Agent node in graph mode.