har.md

HAR recording

--har PATH writes one HTTP Archive (HAR) 1.2 sidecar while the normal response continues through the usual output pipeline:

fetch --har request.har https://example.com
fetch --har request.har -o response.bin https://example.com/download

The file records the final effective exchange after redirects, retries, and Digest authentication. Request and response headers retain duplicate values. Timing fields that are not available are -1.

Request and response bodies are captured through a bounded streaming tee. Each body is limited to 16 MiB. UTF-8 text is stored as text; binary data is stored as base64 while within the limit. If a body is larger, the captured payload is omitted entirely and the HAR contains:

Body omitted by fetch because it exceeds the 16 MiB HAR capture limit

Unary gRPC calls are supported when a local schema or successful reflection positively identifies the method as unary. Framed non-text bodies are base64 encoded and trailers and gRPC status remain in the exchange. Schema-less or unresolved gRPC methods are rejected for HAR because their cardinality is not known. WebSocket sessions, DNS/TLS inspection, gRPC list/describe, dry runs, and unbounded gRPC streams do not support HAR recording.

Safety and sensitivity

HAR files are intentionally faithful reproductions. They can contain Authorization headers, cookies, session tokens, request bodies, and response bodies. Store them with private permissions, review them before sharing, and do not commit them to a repository.

The HAR destination cannot be - and cannot be the response output path. The file is written to an exclusive temporary file and committed atomically. Use --clobber to replace an existing regular file. Symlink destinations are rejected.