Skip to main content
Use this page when a user asks Veryfront Agent to create a scheduled agent, schedule an existing agent, or make an agent run on a cadence.

​
Agent-facing flow

  1. Inspect existing agents with list_agents and existing schedules with list_schedules.
  2. If the target agent does not exist, follow Create an agent first, then return to this flow in the same run.
  3. Collect only missing schedule requirements. Required fields are the agent ID, schedule expression, timezone, and run prompt.
  4. If an active schedule already targets the same agent, reuse it unless the user asked to change the cadence, timezone, or prompt; use update_schedule only for that change branch.
  5. Otherwise create a schedule resource whose target is the agent using create_schedule.
  6. Verify the schedule creation result or read the schedule back with list_schedules.

​
Guardrails

  • Use conversation_mode: "create_new" for recurring autonomous agent schedules unless the user explicitly asks to reuse an existing conversation. See Conversation modes for what each mode gives up.
  • Do not create duplicate schedules for the same target agent and cadence.
  • Do not emit progress-only assistant text between create_agent and create_schedule. After create_agent succeeds, the next assistant action must be the create_schedule tool call unless no schedule creation surface exists.

​
Schedule target

Agent schedules use a typed schedule target. The target is the agent itself, not a wrapper task or workflow:
Sending conversation_mode or conversation_id on a task or workflow target is a validation error, not a default — the fields exist only for agent targets.

​
Conversation modes

conversation_mode decides whether each occurrence gets a hosted conversation. It is a durability and observability choice, not a correctness one — a delegating agent works under any mode. create_new stays the recommendation for recurring schedules because it is what makes a past run inspectable in Studio. Under none you still get the run and its events, but no conversation to open. Authoring the same thing from source uses the camelCase spelling on the target, with the run prompt in agentMessage:
Requires veryfront 0.1.1239 or later. Earlier versions reject both conversationMode on the target and agentMessage, and the schedule file fails to load.

​
Tool selection

  • Use list_agents before creating or choosing the target agent.
  • Use list_schedules before create_schedule and again when verification needs a follow-up read.
  • Use create_schedule for new schedules.
  • Use update_schedule only when a matching schedule already exists and the user asked to change it.

​
Example pattern

Use list_agents, list_schedules, then create_schedule for scheduling agents. When the target agent had to be created first, use the created agent ID as target.id and call create_schedule before any status or final response.

​
Output shape

Return a short implementation summary with the created or reused agent ID, created schedule ID, schedule expression, timezone, conversation mode, and verification result. If no schedule creation surface is available in the run, say that the agent was created or found but the schedule could not be attached; do not present the work as complete.