Template Syntax

Four nodes support {{fieldName}} interpolation to inject record data into prompts, URLs, bodies, and tool arguments. The syntax is node-specific — use the wrong form and the placeholder resolves to an empty value silently.

Syntax by Node

LLM and Agent — {{fieldName}}

LLM and Agent nodes resolve {{fieldName}} against top-level fields of the current record in their system prompt and user/initial message fields.

{{company_name}}
{{email}}
{{lead_score}}

Missing fields resolve to an empty string.

MCP — {{row.fieldName}}

The MCP node builds its tool arguments from a JSON template. Placeholders inside that template must use the row. prefix:

{
  "domain": "{{row.domain}}",
  "industry": "{{row.industry}}"
}

Using {{fieldName}} (without row.) inside MCP JSON arguments will not be replaced.

HTTP — {{fieldName}} in URL and body only

The HTTP node resolves {{fieldName}} in the URL template and the request body. Headers are not interpolated — header values are treated as static strings.

Both URL and body interpolation match only \w+ identifiers (letters, digits, underscores). Fields with hyphens or other non-word characters must be renamed upstream with a Transform node before use. URL values are URL-encoded automatically; body values are escaped for safe insertion.

Where Template Syntax Works

RAG Query does not use template interpolation for its search query — you configure a queryColumn to select which record field contains the query text. Its promptTemplate field uses the named variables {{context}} and {{question}}, which are filled by the executor, not by record fields.

Examples

LLM prompt

Write a 2-sentence description for {{company_name}}, a company in {{industry}}.

HTTP URL

https://api.example.com/companies/{{company_id}}/details

MCP tool arguments

{
  "name": "{{row.company_name}}",
  "region": "{{row.region}}"
}

Behavior

• Missing fields: Placeholder replaced with empty string (LLM/Agent/HTTP) or null (MCP JSON).
• Non-string values: Numbers and booleans are converted to their string representation.
• Nested paths: Not supported — only top-level record fields can be referenced. Use a Transform node to flatten nested data before interpolation.