Milestone Start - Nanonets Agents Platform

Starts a new step of the agent’s work, with a name the agent chooses, such as “Read the bank statements”. Everything the agent does until the step ends is grouped under that step in the task feed, the same way configured steps look. Display name “Milestone Start”. The agent ends the step with milestone_complete and a one-line summary, such as “Read 3 statements, 214 transactions.” Different tasks can get different steps: the agent decides each time.

Authentication and enablement

No integration. You don’t add it to the tools list. It is turned on automatically, together with milestone_complete, when the agent’s Dynamic steps setting is on (Settings → Stages, or settings.dynamic_steps: true with instruction_format: "lexical"). Agents with configured steps don’t get it.

Inputs

Field	Type	Required	Description
`name`	string	yes	The step’s name, in a few plain words a business user understands. Leading and trailing spaces are trimmed.

Output

{
  "tool_name": "milestone_start",
  "name": "Read the bank statements",
  "started_at": "2026-10-21T10:15:00Z"
}

Limits and side effects

Starting a step while another is open ends the open one first, with no summary, so no step is left open.
The step is recorded in the task’s step progress and shown as a step header in the task feed. Reopening a finished task shows the same steps, names and summaries.
On the LangChain harness, the agent’s Plan card follows the steps. Starting a step marks the matching plan item in progress, adding it if it is missing. Finishing the step ticks the item off.
There is no limit on the number of steps in a task.
Free: no charge, and no external call.

Expected errors

name is required: the name was empty.
this agent's steps are configured; it cannot start its own: the agent has configured steps, or Dynamic steps is off.

​Authentication and enablement

​Inputs

​Output

​Limits and side effects

​Expected errors

Authentication and enablement

Inputs

Output

Limits and side effects

Expected errors