---
title: "WARC archives"
description: "Capture pages to WARC, extract from archives at high volume, and replay them."
---

```sh
webfold capture <url> -o out.warc.gz [--backend native] [--tlp RED|AMBER|AMBER+STRICT|GREEN] [--retention-days N] [--no-gzip]
webfold warc   <file.warc|file.warc.gz> [--threads n] [--max n]
webfold dump   <warc-file> <out-file> [--max n]
webfold replay <file.warc|file.warc.gz> [--out replay.html] [--max n] [--full-html]
```

| Command | Does |
| --- | --- |
| `capture` | Fetch one URL through the native backend into a WARC and a `.cdx` index (warcinfo, request, response and metadata records). |
| `warc` | Run extraction over a WARC or WARC.GZ stream and print a summary. Use `--threads` to parallelise a raw `.warc`; a `.warc.gz` is streamed on one thread and ignores `--threads`. |
| `dump` | Write the response pages out of a WARC so another tool can run over the identical corpus. |
| `replay` | Join each record's raw HTML with its canonical extraction into a single-file HTML audit artefact. |

:::note
Tor is not available in the main binary. Its adapter is an unvalidated opt-in in the separate `integrations/` package.
:::

Throughput depends heavily on hardware, thread count and corpus. See [Benchmarks](/docs/reference/benchmarks) for what has been measured and under which protocol.