Get Task State - Nanonets Agents Platform

Reads another task’s task state: the output schema of that task’s agent, filled so far from the task’s tool outputs by the agent’s saved state mappings, with reviewer corrections applied. Display name “Get Task State”. Off by default. Use it when one task reports on another, for example a summary agent that a hook starts with the source task’s id, and that turns the source task’s journal into the payload a customer system expects.

Authentication and enablement

No integration. Off by default; enable it on the agent’s tools list. Read-only. The task being read must belong to an agent that has an output schema and task state turned on, with its mappings saved. Otherwise the tool returns configured: false and no state.

Inputs

Field	Type	Required	Description
`task_id`	string (UUID)	yes	The task to read. It must be in the same workspace as the running task.

A hook-started task carries the source task’s id in its source metadata; read it with the task_metadata tool.

Output

{
  "task_id": "8c1d…e42a",
  "agent_group_id": "263974f9…",
  "title": "Payroll accrual 05.2026",
  "status": "waiting_for_input",
  "state_enabled": true,
  "configured": true,
  "state": {
    "journal": { "entry_id": "PAY-05-2026-001", "currency": "CHF", "amount": 477500 },
    "lines": [ { "gl_account": "620100", "amount": 125400 } ]
  },
  "unmapped_tools": ["output_file_index"]
}

state: the schema’s fields, with null for a field nothing has filled yet, and every table row carrying every column. It is null when configured is false.
state_enabled: whether task state is turned on for the deployment.
configured: whether the task’s agent has an output schema and task state turned on.
unmapped_tools: tools that ran in the task but have no saved mapping, so their output is not in state yet. Map them in the agent’s State settings.
status and title are the task’s own, so a summary can say whether the task is still running, paused or finished.

Limits and side effects

No side effects. The state is read with the saved mappings only: the tool never generates a mapping and never calls a model, so the same task reads the same way every time.
Same workspace only. A task in another workspace returns the same error as a task that does not exist.
Files a task produced are not part of the state; only the fields the schema defines are.

Expected errors

task <id> not found: no task with that id in this workspace.
task_id must be a task UUID: the input is not a UUID (for example a journal entry number).
no workspace in context: called outside a running task.
task state is unavailable: misconfigured deployment.

​Authentication and enablement

​Inputs

​Output

​Limits and side effects

​Expected errors

Authentication and enablement

Inputs

Output

Limits and side effects

Expected errors