Skip to main content

Tool Nodes

Tool nodes let your workflow interact with external systems. Call HTTP APIs, connect to MCP servers, and integrate with third-party services directly from your workflow.
Security: Tool credentials are encrypted at rest using AES-256-GCM and are never exposed in API responses. Credentials are sent once during creation and cannot be retrieved afterward.

tool.http_request

Call an external HTTP endpoint. The HTTP Request node makes outbound API calls to external services. Use it to integrate with your CRM, ticketing system, payment provider, or any REST API.

Configuration

OptionTypeRequiredDescription
methodstringYesHTTP method: GET, POST, PUT, DELETE
urlstringYesThe endpoint URL. Supports context variable substitution
headersobjectNoRequest headers as key-value pairs (e.g., Authorization, Content-Type)
bodystringNoRequest body template (for POST/PUT). Supports context variables
timeoutnumberNoRequest timeout in milliseconds (default: 10000)
credentialIdstringNoReference to a stored credential for authentication

URL and Body Variables

Use context variables in URLs and request bodies:
# URL with variable
https://api.example.com/orders/{{conversation.fields.order_id}}

# JSON body with variables
{
  "customer_phone": "{{contact.phone}}",
  "customer_name": "{{contact.name}}",
  "inquiry_type": "{{router.category}}"
}

Security Protections

The HTTP Request node includes built-in security measures:
  • SSRF protection: Private IP addresses and internal network ranges are blocked
  • URL allowlists: Admins can configure which domains are permitted
  • Encrypted credentials: API keys and tokens are stored encrypted, never logged or exposed
  • Request logging: All outbound requests are logged in run steps (with sensitive headers redacted)

Outputs

OutputTypeDescription
response.statusnumberHTTP status code (200, 404, 500, etc.)
response.bodyobject/stringThe parsed response body
response.headersobjectResponse headers

Connections

  • Inputs: Single input handle from the previous node
  • Outputs: Single output handle passing the response to the next node

Example Use Case

After a customer asks about their order status, use an HTTP Request node to call your e-commerce API: GET https://api.myshop.com/orders/{{conversation.fields.order_id}}/status. Pass the response to an AI Respond node that can formulate a natural language update for the customer.

tool.mcp_call

Call a tool on an MCP (Model Context Protocol) server. The MCP Call node connects to MCP-compatible servers, enabling advanced integrations with services that implement the Model Context Protocol. MCP provides a standardized way for AI applications to interact with external tools and data sources.

Configuration

OptionTypeRequiredDescription
serverUrlstringYesThe MCP server URL
toolNamestringYesThe name of the tool to invoke on the MCP server
parametersobjectNoTool-specific parameters as key-value pairs. Supports context variables
credentialIdstringNoReference to a stored credential for MCP server authentication

Outputs

OutputTypeDescription
resultobjectThe tool execution result returned by the MCP server
isErrorbooleanWhether the tool execution resulted in an error

Connections

  • Inputs: Single input handle from the previous node
  • Outputs: Single output handle passing the MCP tool result to the next node

Example Use Case

Connect to an MCP server that provides database query tools. Call a search_knowledge_base tool with the customer’s question as input, then pass the results to an AI Respond node to generate an answer grounded in your proprietary data.
MCP is an advanced integration pattern. Most workflows will use tool.http_request for standard API integrations. Use tool.mcp_call when integrating with services that specifically implement the Model Context Protocol.

Managing Tool Credentials

Tool credentials are managed in Settings > Tools within your workspace:
  1. Create a tool definition with the endpoint URL and method
  2. Add credentials (API key, bearer token, or custom headers) — these are encrypted immediately
  3. Test the connection using the built-in test button
  4. Reference the tool in your workflow nodes by selecting it from the tool picker in the inspector panel
Credentials are:
  • Encrypted at rest with AES-256-GCM
  • Sent once during creation, never returned in API responses
  • Scoped to your workspace (tenant isolation enforced)
  • Visible only to Admin role users