Install the ReyGrid community node in n8n to chat with your RAG agents, run connected AI tools automatically, and cancel tool calls.
The ReyGrid N8N node is published on npm as the package n8n-nodes-reygrid. It is a community node for n8n. It adds a single ReyGrid node to your workflows that calls the ReyGrid REST API.
Sends a message to an agent (POST /agents/{agentId}/chat). AI tools connected to the node are executed automatically until the agent returns its final answer.
In n8n, go to Settings → Community nodes and click Install a community node.
Settings → Community nodes
2
Enter the package name
In npm Package Name, enter:
n8n-nodes-reygrid
Tick I understand the risks of installing unverified code from a public source, then click Install.
Install community nodes dialog
3
Find the node
When the installation finishes, open a workflow, click +, and search for ReyGrid.
Community nodes are installed by an n8n instance owner or admin. If you don’t see Community nodes in Settings, ask the person who manages your n8n instance.
Install from the command line instead
On a self-hosted instance you can install the package into n8n’s custom nodes folder and restart n8n:
cd ~/.n8n/nodesnpm install n8n-nodes-reygrid
The folder location can differ depending on how you run n8n (for example in Docker).
The example below receives a request on a Webhook node and sends it to a ReyGrid agent. A Calculator tool is connected to the node’s Tools input, so the agent can use it.
ReyGrid node with a Calculator tool connected to its Tools input
Stable user identifier used as a persistent conversation. When set, Conversation ID is ignored.
Conversation ID
no
An existing conversation. Ignored when User ID is set. Leave both empty to start a new conversation.
Both identifiers behave as described in Persistent conversations. If you use only Conversation ID, the node follows the conversation ID returned by the API for the remaining rounds of the same run.Options
Option
Default
Description
Include Tool Calls in Output
off
Adds a toolCalls array (name, arguments, output) describing every tool executed during the run.
Max Tool Iterations
10
Maximum number of tool-execution rounds. When exceeded, pending tool calls are cancelled and the node fails.
On Tool Error
Send Error to Agent
Send Error to Agent returns the error text as the tool result so the agent can recover. Cancel Tool Calls and Stop cancels the pending tool calls through the API and returns the last response.
Tools connected to the Tools input are sent to the agent as function tools. The node then repeats the following until the agent has no more tool calls:
Send the message (or the previous tool results) to the agent.
If the response contains tool calls, run the matching connected tools.
Send the tool outputs back as results and continue.
Tool calls are read from output.tool_calls, output.reply.tool_calls, or tool_calls in the response. If the agent asks for a tool that isn’t connected, or a tool fails, the behavior follows On Tool Error. Connected tool names must be unique.
The OpenAPI specification doesn’t describe where a chat response carries tool calls. This page documents what the node does, not a guaranteed API contract. See Tool calling.
API failures are shown as n8n node errors. The message comes from the API error body (code: message for most errors, or the error text for 401). Turn on Continue On Fail in the node settings to receive { "error": "…" } for a failed item instead of stopping the workflow.See Errors for the status codes.