---
title: deleteRuns
description: Delete the stored data of finished workflow runs in one call.
docs_index: /llms.txt
lastUpdated: 2026-11-19
type: reference
summary: Use deleteRuns to delete the inputs, outputs, streams, and hook payloads of up to 100 finished runs.
related:
  - /docs/observability/retention
  - /docs/api-reference/workflow-api/get-run
---

> For an index of all documentation, see [/llms.txt](/llms.txt).

Deletes the stored data of up to 100 finished runs: their inputs, outputs and errors, their steps' and events' payloads, their hook payloads, and their stream contents. Metadata is not deleted: the runs stay listed, and so do their steps, events and hooks, with their IDs, names, statuses, timestamps and attributes. Only the payloads go, reading back as expired. This cannot be undone. See [Data retention](/docs/observability/retention#deleting-a-finished-runs-data).

For a single run, call [`run.delete()`](/docs/api-reference/workflow-api/get-run#delete-a-finished-runs-data) instead.

```typescript lineNumbers
import { deleteRuns } from "workflow/api";

const { summary, results } = await deleteRuns([runIdA, runIdB]); // [!code highlight]
```

`deleteRuns()` never cancels a run: a run that hasn't finished is answered `skipped` and keeps running. Cancel it first to delete its data now.

## API signature

### Parameters

### Returns

A summary that counts each outcome, and one result per run in the order of `runIds`. Each result has one of these outcomes:

| Outcome           | Meaning                                                                                           |
| ----------------- | ------------------------------------------------------------------------------------------------- |
| `deleted`         | The run's data was deleted.                                                                       |
| `already_deleted` | The run's data was already gone, from an earlier deletion or `experimental_retention: 0`.         |
| `skipped`         | The run hasn't finished, so nothing was deleted. `status` is the run's status.                    |
| `not_created`     | The run was cancelled before it was created. What it held is deleted.                             |
| `not_found`       | No run has this ID.                                                                               |
| `failed`          | The deletion didn't complete. `code` says why, and `retryable` says whether retrying can succeed. |

## Retries

Deletion is idempotent, so retrying is always safe: a run that was already handled answers `already_deleted`.

- Inside a step, runs answered `failed` with `retryable: true` are sent again once, on their own, a second later. If any still fail, or the World rate-limits a request, it throws a [`RetryableError`](/docs/api-reference/workflow/retryable-error) so the step retries the whole batch, after the World's requested delay, or 10 seconds when it gives none.
- Outside a step, `failed` results are returned as they are, so retry the runs answered `failed` with `retryable: true` yourself. A rate limit throws the World's `ThrottleError`, whose `retryAfter` is the number of seconds to wait.

## Errors

`deleteRuns()` throws when `runIds` is empty, contains duplicates, or has more than 100 IDs, and when the configured World doesn't support deleting run data. Send more runs in batches of 100. The World enforces the same limit.

---

For a semantic overview of all documentation, see [/sitemap.md](/sitemap.md)

For an index of all available documentation, see [/llms.txt](/llms.txt)

For agent-facing discovery, including API and MCP surfaces, see [/agents.md](/agents.md)