Destination Properties by Connector

Reference for all properties available in the to block when writing to each connector type. Every to block shares a set of universal attributes; the sections below document what target, operation, query, and params mean for each connector.

---

Universal Attributes

Available on every to block regardless of connector type:

Attribute	Type	Default	Description
connector	string	required	Target connector name
target	string	—	Resource identifier (meaning varies per connector — see below)
operation	string	auto	Override operation type (meaning varies per connector)
format	string	json	Output format: json, xml
query	string	—	Raw SQL with named parameters (:name, :id) — resolved from transformed payload
query_filter	map	—	NoSQL filter document (MongoDB)
update	map	—	NoSQL update document (MongoDB $set, $inc, etc.)
params	map	—	Extra connector-specific parameters (CEL expressions)
when	string	—	CEL condition — only write if true. Context: input, output
parallel	bool	true	In multi-to, run this destination in parallel
required	bool	false	In multi-to, this destination decides the outcome: it runs before the others, and if it fails the flow fails (retry, DLQ, on_error) and the others do not run
transform	block	—	Per-destination CEL transform (overrides flow-level transform)

Data mapping

to.target        → connector.Data.Target
to.operation     → connector.Data.Operation
to.query         → connector.Data.RawSQL
to.query_filter  → connector.Data.Filters
to.update        → connector.Data.Update
to.params        → connector.Data.Params
transformed data → connector.Data.Payload

How a params value is read

Each value is resolved the same way a query_filter value is, and the rule is worth knowing because two of the three forms look like expressions:

Written as	Read as
"input.sheet"	The expression, evaluated against the message
":sku"	The path parameter of that name
true, 42, "csv"	The literal, as written

A string is only evaluated when it mentions input. — so append = "1 == 1" is the five-character string, not true. Write literals as literals: append = true.

---

Database (SQLite, PostgreSQL, MySQL)

Property	Value
target	Table name (e.g., "users", "orders")
operation	INSERT (default for POST), UPDATE, DELETE
query	Raw SQL with named parameters (:name, :email). Resolved from payload
params	Not used (named params come from payload)

Named parameters in query

When using query, named parameters like :name are replaced with the corresponding field from the transformed payload:

transform {
  number = "input.payload.associateNumber"
  name   = "input.payload.name"
  emails = "input.payload.emails.join(',')"
}

to {
  connector = "magento_db"
  target    = "sales_associate"
  query     = "INSERT INTO sales_associate (number, name, emails) VALUES (:number, :name, :emails) ON DUPLICATE KEY UPDATE name = :name, emails = :emails"
}

A placeholder is only recognised where the statement is code. Colons inside a comment, a string literal or a quoted identifier are left exactly as written, and so are Postgres casts:

-- ratio:sku is a comment, not a parameter
SELECT id, 'a:b' AS label, :sku::text
FROM catalog
WHERE sku = :sku          -- the item's parent may be promoted

Only the two :sku in the SELECT and WHERE bind; ratio:sku, 'a:b', and the ::text cast are untouched. Comments are read per driver, so MySQL's # and its rule that -- needs whitespace after it both apply where they should.

Fixed in 3.3.0

Before 3.3.0 the binder knew about string literals and nothing else, so an apostrophe in a comment — -- the item's parent — opened a literal that never closed, and every placeholder after it reached the driver unbound. The statement failed with missing named argument, naming the parameter rather than the comment. In the other direction, -- ratio:sku was bound as if it were a placeholder and consumed one of the statement's arguments. mycel validate does not execute SQL, so neither showed up there.

A placeholder with no matching value is left as written rather than bound to nothing, so the driver rejects the statement you wrote instead of one with an argument silently missing.

Binding a set — IN (:name)

A list bound inside an IN (...) is expanded into as many placeholders as it has members:

SELECT * FROM order_items WHERE order_id IN (:order_ids)

with order_ids holding three values becomes IN (?, ?, ?) and three arguments — or IN ($2, $3, $4) on Postgres, numbered in sequence with everything around it. NOT IN works the same way. A string is not a list: IN (:name) with a name in it means one name.

Two shapes are refused rather than guessed, and both name the parameter:

What	Why
An empty list	IN () is not valid SQL in MySQL, Postgres or SQLite. There is no expansion right for both directions — IN (NULL) matches nothing, which is what an empty set means, but NOT IN (NULL) also matches nothing, which is its opposite. Guard the statement with when on the step instead
A list where a scalar belongs — WHERE id = :ids	Expanding gives id = ?, ?, ?, which the driver rejects with a position in the statement rather than the name of the parameter

The guard for the empty case:

step "items" {
  connector = "db"
  query     = "SELECT * FROM order_items WHERE order_id IN (:order_ids)"
  when      = "size(step.orders ?? []) > 0"
  params    = { order_ids = "pluck(step.orders, 'id')" }
}

Fixed in 3.4.0

Before 3.4.0 a list was bound whole to a single placeholder, and database/sql refuses that on every driver: unsupported type []interface {}, a slice of interface. examples/steps shipped a flow doing exactly this, and no curl in its README reached that flow, so nothing ever ran it.

Standard operations (no query)

Without query, the operation is inferred from the HTTP method or set explicitly:

# INSERT — payload fields become columns
to {
  connector = "db"
  target    = "users"
}

# UPDATE — filters from URL params, payload = SET clause
to {
  connector = "db"
  target    = "users"
  operation = "UPDATE"
}

# DELETE — filters from URL params
to {
  connector = "db"
  target    = "users"
  operation = "DELETE"
}

PostgreSQL specifics

• INSERT ... RETURNING * returns the full created row (including id, created_at, etc.)

---

MongoDB

Property	Value
target	Collection name (e.g., "users")
operation	INSERT_ONE, INSERT_MANY, UPDATE_ONE, UPDATE_MANY, DELETE_ONE, DELETE_MANY, REPLACE_ONE
query_filter	MongoDB filter document (WHERE equivalent)
update	MongoDB update document ($set, $inc, $push, etc.)
params	{ upsert = true }, { documents = [...] } for INSERT_MANY

to {
  connector    = "mongodb"
  target       = "orders"
  operation    = "UPDATE_ONE"
  query_filter = { order_id = "input.order_id" }
  update       = { "$set" = { status = "completed", updated_at = "now()" } }
}

---

Message Queues (RabbitMQ, Kafka, Redis Pub/Sub, MQTT)

Property	RabbitMQ	Kafka	Redis Pub/Sub	MQTT
target	Routing key	Topic	Channel	Topic
operation	PUBLISH (implicit)	PUBLISH	PUBLISH	PUBLISH
params	{ exchange = "..." }	—	—	{ qos = 1, retain = true }

# RabbitMQ
to {
  connector = "rabbit"
  target    = "order.created"
}

# Kafka
to {
  connector = "kafka"
  target    = "orders"
}

# MQTT with QoS
to {
  connector = "mqtt"
  target    = "sensors/temperature"
  params    = { qos = 1, retain = true }
}

The transformed payload becomes the message body (JSON).

---

HTTP Client

Property	Value
target	Endpoint path (e.g., "/api/users", "POST /api/notify")
operation	HTTP method override: GET, POST, PUT, PATCH, DELETE

to {
  connector = "external_api"
  target    = "/webhooks/order-created"
  operation = "POST"
}

The transformed payload becomes the request body. For read verbs, which carry no body, the payload's fields become query string parameters instead.

---

GraphQL Client

Property	Value
target	Full GraphQL query/mutation string
operation	Not used (embedded in the query string)

to {
  connector = "graphql_api"
  target    = <<-EOF
    mutation CreateUser($input: UserInput!) {
      createUser(input: $input) { id name email }
    }
  EOF
}

The transformed payload becomes the GraphQL variables.

---

gRPC Client

Property	Value
target	RPC method name (e.g., "CreateUser", "users.UserService/CreateUser")
operation	Alternative to target for the method name

to {
  connector = "grpc_service"
  target    = "CreateUser"
}

The transformed payload becomes the protobuf message fields.

---

SOAP Client

Property	Value
target	SOAP operation name (e.g., "CreateItem", "GetOrder")
operation	Alternative to target for the operation name

to {
  connector = "soap_service"
  target    = "CreateItem"
}

The transformed payload becomes the SOAP body parameters.

---

File

Property	Value
target	File path (relative to connector base_path)
operation	WRITE (default), DELETE, COPY, MOVE
params	{ format = "csv" }, { append = true }, { sheet = "Data" } (Excel)

# Write JSON
to {
  connector = "files"
  target    = "output/report.json"
}

# Append to CSV
to {
  connector = "files"
  target    = "logs/access.csv"
  params    = { format = "csv", append = true }
}

# Write to Excel sheet
to {
  connector = "files"
  target    = "reports/monthly.xlsx"
  params    = { sheet = "March" }
}

Format is auto-detected from file extension. The transformed payload becomes the file content.

---

S3

Property	Value
target	S3 object key (e.g., "uploads/document.pdf")
operation	PUT (default), DELETE, COPY
params	content, content_type, storage_class, acl, metadata

to {
  connector = "s3"
  target    = "'uploads/' + input.user_id + '/avatar.png'"
  params    = {
    content       = "output._binary"
    content_type  = "'image/png'"
    storage_class = "'STANDARD'"
  }
}

---

Exec

Property	Value
target	Command to execute
operation	Not used
params	{ args = [...] }, { stdin = "..." }

to {
  connector = "exec"
  target    = "convert"
  params    = { args = ["-resize", "800x600", "input.file_path", "output.jpg"] }
}

---

Elasticsearch

Property	Value
target	Index name (e.g., "products")
operation	index (default), update, delete, bulk

to {
  connector = "search"
  target    = "products"
  operation = "index"
}

The transformed payload becomes the document to index. Returns _id, status.

---

PDF

Property	Value
target	Template path fallback (if not in connector config or payload)
operation	generate (returns binary for HTTP) or save (writes file)

to {
  connector = "invoice_pdf"
  operation = "generate"
}

Template is resolved: payload template field > connector config template > target fallback. All other payload fields become template variables ({{.field_name}}). Special payload fields: filename (for Content-Disposition).

---

WebSocket

Property	Value
target	Room name (for send_to_room)
operation	broadcast, send_to_room, send_to_user

# Broadcast to all clients
to {
  connector = "ws"
  operation = "broadcast"
}

# Send to specific room
to {
  connector = "ws"
  operation = "send_to_room"
  target    = "order-updates"
}

# Send to specific user (user_id from payload)
to {
  connector = "ws"
  operation = "send_to_user"
}

For send_to_user, the payload or filters must include user_id.

---

SSE (Server-Sent Events)

Property	Value
target	Room name
operation	broadcast, send_to_room

to {
  connector = "sse"
  operation = "send_to_room"
  target    = "dashboard"
}

---

TCP Client

Property	Value
target	Connection identifier
operation	SEND (implicit)

The payload is serialized according to the connector's protocol setting (json, msgpack, nestjs).

---

Notification Connectors

Notification connectors receive the transformed payload as the message. The payload fields map to the notification's properties.

Email

Payload field	Type	Description
to	array	[{email, name}] recipients
subject	string	Email subject
text_body	string	Plain text body
html_body	string	HTML body
template	string	Override connector-level template path
template_data	map	Variables for template rendering
cc, bcc	array	CC/BCC recipients
attachments	array	[{filename, content, content_type}]

to {
  connector = "email_smtp"
  operation = "send"
}

Template resolution: payload template > connector config template.

Slack

Payload field	Type	Description
text	string	Message text
channel	string	Channel name or ID
blocks	array	Slack Block Kit blocks
thread_ts	string	For threaded replies

Discord

Payload field	Type	Description
content	string	Message text
channel_id	string	Channel ID
embeds	array	Discord embed objects

SMS

Payload field	Type	Description
to	string	Phone number (E.164)
message	string	SMS body

Push

One of token, tokens, topic or condition says who receives it.

Payload field	Type	Description
token	string	The device to send to
tokens	list	Several devices at once
topic	string	Everyone subscribed to a topic
condition	string	Everyone matching a topic condition
title	string	Notification title
body	string	Notification body
data	map	Custom data payload
priority	string	normal or high
collapse_key	string	Replaces an earlier undelivered notification with the same key
ttl	int	Seconds the service should keep trying

Webhook

Payload field	Type	Description
url	string	Override connector URL
method	string	HTTP method
headers	map	Extra headers
body	any	Request body

---

Special Response Fields

These fields in the flow result have special meaning when returned through the REST connector:

Field	Type	Recognized by	Purpose
_binary	string (base64)	REST	Serve as binary download
_content_type	string	REST	MIME type for binary response
_filename	string	REST	Content-Disposition filename
http_status_code	int/string	REST, SOAP	Override HTTP status code
grpc_status_code	int/string	gRPC	Override gRPC status code
_response_headers	map	REST (aspects)	Extra HTTP response headers

---

See also: Flows for to block syntax, Configuration Reference for all HCL blocks.