Execute operations
connectors execute runs an action against an entity on a configured connector. Use it for reads, searches, creates, and updates.
Always run connectors describe before executing. The describe response is the source of truth for entity names, action names, action parameters, and fields.
Basic command
airbyte-agent connectors execute --json '{
"workspace": "default",
"name": "GitHub",
"entity": "issues",
"action": "context_store_search",
"select_fields": ["id", "title", "state"],
"params": {
"limit": 10
}
}'
Required input:
id, ornamewithworkspace, to choose the connector.entity, such asissues,contacts, ormessages.action, such ascontext_store_search,list,get,api_search,create, orupdate.
Optional input:
params: action-specific parameters.select_fields: fields to include in the connector response.exclude_fields: fields to remove from the connector response.skip_truncation: disables automatic truncation of long text fields in list and search responses. This maps to--skip-truncationwhen using per-parameter flags.
Choose an action
The actions available on a connector come from connectors describe. Many connectors support some of these actions:
context_store_search: Search or filter the indexed Context Store. Prefer this for reads when indexed data is acceptable.list: Read live data from the source. Use this when you need data that may not be indexed yet.get: Fetch one record by ID.api_search: Use a provider-native search surface, when the connector exposes one.create: Create a record in the source service.update: Update a record in the source service.
Don't guess an action name. If it isn't in the describe output for that connector and entity, don't call it.
Pass action parameters
Action-specific parameters go under params:
airbyte-agent connectors execute --json '{
"workspace": "default",
"name": "GitHub",
"entity": "issues",
"action": "context_store_search",
"select_fields": ["id", "title", "state"],
"params": {
"limit": 20,
"query": {
"filter": {
"equals": {
"state": "open"
}
}
}
}
}'
The exact params shape is connector-, entity-, and action-specific. Read it from connectors describe.
Filter response fields
Use two layers of filtering when you know which fields you need.
API-side filtering: select_fields and exclude_fields
select_fields and exclude_fields are part of the JSON payload and are forwarded to the connector execution API.
Use select_fields to keep only specific fields:
airbyte-agent connectors execute --json '{
"workspace": "default",
"name": "GitHub",
"entity": "issues",
"action": "context_store_search",
"select_fields": ["id", "title", "state"],
"params": {
"limit": 10
}
}'
Use exclude_fields to drop heavy or unnecessary fields:
airbyte-agent connectors execute --json '{
"workspace": "default",
"name": "Slack",
"entity": "messages",
"action": "list",
"exclude_fields": ["blocks", "attachments"],
"params": {
"limit": 10
}
}'
If you pass both, the API gives select_fields priority.
Client-side filtering: --fields
--fields filters the JSON after the API returns it. connectors execute returns {"status":"success","result": ...} on success, so field paths start under result.
For object results, name the object fields you want:
airbyte-agent connectors execute --json @request.json --fields result.id,result.name
For array results, name the row fields you want. The CLI applies the field path to each row:
airbyte-agent connectors execute --json @request.json --fields result.number,result.title,result.state
Paths use dot notation. When a path crosses an array, the remaining path applies to every element. For list-style responses wrapped in {"data": [...]}, the CLI can also broadcast row-level paths:
airbyte-agent connectors list --json '{"workspace": "default"}' --fields id,name
Load JSON from a file
For long payloads, write the request to a file:
{
"workspace": "default",
"name": "GitHub",
"entity": "issues",
"action": "context_store_search",
"select_fields": ["id", "title", "state"],
"params": {
"limit": 10
}
}
Then run:
airbyte-agent connectors execute --json @request.json
@filename is resolved relative to the current working directory.
Pagination
Pagination is connector- and action-specific. Use connectors describe to check whether the action returns a cursor, page token, offset, or provider-specific pagination fields. Pass the next-page value back under params using the parameter name documented by that action.