> ## Documentation Index
> Fetch the complete documentation index at: https://docs.crewship.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Retry

> Re-run a previous run with the same agent + recovered prompt — or with a tweaked prompt — into a new chat or continuing the old one.

# Retry

`cmd/crewship/cmd_retry.go` looks up a prior run, fetches its first user prompt from chat messages, and starts a new run against the same agent. Iteration workflow:

```bash theme={null}
crewship history --status failed
crewship retry <run-id>                         # exact same again
crewship retry <run-id> --new-prompt "..."      # same agent, edited prompt
```

The run lookup uses the `/api/v1/runs?limit=100` list and filters client-side — if the run is older than the most recent 100, the command fails with `run <id> not found in last 100 runs`. Targeted re-runs are the intended use case; for archival lookup, fetch the chat ID some other way and pass `--new-prompt`.

## `crewship retry <run-id>`

| Flag                  | Type   | Default  | Effect                                                                                                                                        |
| --------------------- | ------ | -------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
| `--new-prompt <text>` | string | (unset)  | Override the recovered prompt with this text. Required when the run has no chat (e.g. `run <id> has no chat_id; pass --new-prompt to retry`). |
| `--continue`          | bool   | `false`  | Append to the original chat instead of starting a new one. Requires the original to have `chat_id`.                                           |
| `-q`, `--quiet`       | bool   | `false`  | Only output agent text — suppress meta lines.                                                                                                 |
| `--no-stream`         | bool   | `false`  | Wait for completion then print, instead of streaming.                                                                                         |
| `--markdown`          | bool   | (config) | Force markdown ANSI styling on.                                                                                                               |
| `--no-markdown`       | bool   | (config) | Force markdown styling off.                                                                                                                   |
| `--save <path>`       | string | (unset)  | Tee the agent's text response (no ANSI) to file.                                                                                              |

### Examples

```bash theme={null}
crewship retry r_abc                            # same agent, same prompt, new chat
crewship retry r_abc --continue                 # append to the same chat
crewship retry r_abc --new-prompt "be more specific"
crewship retry r_abc --quiet --save out.md
crewship retry r_abc --no-stream
```

### Common errors

* **`run <id> not found in last 100 runs`** — run is older than the recent window; pass a `--new-prompt` and use a different command.
* **`run <id> has no chat_id; pass --new-prompt to retry`** — the original run never produced a chat (rare; e.g. pre-chat trace runs).
* **`could not recover original prompt for run <id>; pass --new-prompt`** — the chat exists but no user-role message was found.
* **`--continue requires the original run to have a chat_id`** — combine `--continue` with `--new-prompt` against a runnable chat instead.

## See also

* [`crewship history`](/cli/history) — find the run-id to retry.
* [`crewship copy-prompt <run-id>`](/cli/copy-prompt) — recover the prompt without starting a new run.
* [`crewship resume <run-id>`](/cli/resume) — continue the existing chat interactively.