kind'llm' | 'browser'
What executes the agent's runs. llm agents run their instructions in a sandbox. browser agents run a fixed automation against a web portal and import what it exports; they have no instructions, are set up by End Close, and cannot be created or converted through the API.
namestring required
Unique within the environment
descriptionstring nullable
instructionsstring required
The standing prompt every run executes. Empty for browser agents
modelstring nullable
The model the agent runs on — resolved to the platform default when not set explicitly. Null for browser agents, which run on no model
cadence'manual' | 'hourly' | 'daily' | 'weekdays' | 'weekly'
manual agents only run when triggered (POST /agents/{id}/runs)
hourinteger nullable
Required for daily, weekdays, and weekly cadences
minuteinteger nullable
Required for every cadence except manual
day_of_weekinteger nullable
0 = Sunday. Required for weekly cadence
timezonestring
IANA timezone the schedule is evaluated in
activeboolean
Inactive agents are never scheduled but can still be run on demand
handles_file_dropsboolean
Whether this agent processes files dropped in the app. At most one active agent per environment can have this enabled.
endclose_writesboolean
Whether this agent may change End Close data: matches, exceptions and findings, marking records overdue, and starting other agents. When false its runs can only read, and a write from the run is refused with 403.
delivery_enabledboolean
Email output files matching delivery_file_pattern to delivery_recipients after each completed run
delivery_recipientsstring[]
delivery_file_patternstring
Case-insensitive glob selecting which output files are delivered
next_run_atstring date-time nullable
last_run_atstring date-time nullable
created_atstring date-time
updated_atstring date-time