Complete reference for all fetch command-line options. This file is embedded
in the Go binary and is displayed by fetch -v --help.
See also the documentation index, limits and safety, and the dedicated article, HAR, updates, and Agent Skill guides.
fetch [OPTIONS] [URL]
When no scheme is provided, fetch defaults to HTTPS for hostnames. localhost and all IP literals default to HTTP. Dry-run output includes the normalized absolute URL. If a schemeless HTTPS connection fails during setup, fetch suggests the equivalent http:// URL for plaintext services.
fetch example.com # https://example.com
fetch localhost:3000 # http://localhost:3000
fetch http://example.com # Force HTTPA timeout value of 0 disables that deadline. --timeout 0 removes the
request deadline, and --connect-timeout 0 removes the connection setup
deadline.
Specify the HTTP method. Default: GET, or POST when a request-body flag is used without an explicit method.
Alias: -X
fetch -m POST example.com
fetch -X DELETE example.com/resource/123Set custom headers. Can be used multiple times.
fetch -H "Authorization: Bearer token" example.com
fetch -H "X-Custom: value" -H "Accept: application/json" example.comAppend query parameters to the URL in input order. Duplicate parameters are preserved. Can be used multiple times.
fetch -q page=1 -q limit=50 example.comBody options are mutually exclusive - only one can be used per request.
Send a raw request body. Content-Type is auto-detected when using file references.
fetch -d 'Hello, world!' -m PUT example.com
fetch -d @data.txt -m PUT example.com
fetch -d @- -m PUT example.com < data.txtSend a JSON request body. Sets Content-Type: application/json.
fetch -j '{"hello": "world"}' -m POST example.com
fetch -j @data.json -m POST example.comSend an XML request body. Sets Content-Type: application/xml.
fetch -x '<Tag>value</Tag>' -m PUT example.com
fetch -x @data.xml -m PUT example.comSend a URL-encoded form body. Can be used multiple times.
fetch -f username=john -f password=secret -m POST example.com/loginSend a multipart form body. Use @ prefix for file uploads. Can be used multiple times.
fetch -F hello=world -F file=@document.pdf -m POST example.com/uploadOpen an editor to modify the request body before sending. Uses VISUAL or EDITOR environment variables.
fetch --edit -m PUT example.comAuthentication options are mutually exclusive.
HTTP Basic Authentication.
fetch --basic username:password example.comHTTP Digest Authentication. Uses a bounded RFC 7616 challenge-response handshake. Supports MD5, MD5-sess, SHA-256, SHA-256-sess, SHA-512-256, and SHA-512-256-sess, with no qop or qop=auth challenges.
fetch --digest username:password example.comHTTP Bearer Token Authentication.
fetch --bearer mysecrettoken example.comSign requests with AWS Signature V4. Requires AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY environment variables. If AWS_SESSION_TOKEN is set, it is sent and signed as x-amz-security-token.
fetch --aws-sigv4 us-east-1/s3 s3.amazonaws.com/bucket/keyClient certificate file for mTLS. PEM format.
fetch --cert client.crt --key client.key example.comClient private key file for mTLS. Required if --cert is a certificate-only file.
fetch --cert client.crt --key client.key example.comWrite response body to a file. Use - for stdout (bypasses binary detection).
fetch -o response.json example.com/api/data
fetch -o - example.com/file.bin > output.binWrite response body to current directory using the filename from the URL.
Alias: --output-current-dir
fetch -O example.com/path/to/file.txt # Creates ./file.txtUse filename from Content-Disposition header. Requires -O. The extended
filename* parameter is preferred. Unsafe or malformed names fall back to the
URL filename with a warning.
fetch -O -J example.com/downloadOverwrite existing output file (default behavior is to fail if file exists).
fetch -o output.json --clobber example.com/dataCopy the response body to the system clipboard. The response is still printed to stdout normally. Works with all response types including streaming responses (SSE, NDJSON, gRPC). Responses exceeding 1 MiB are not copied to the clipboard.
Requires a clipboard command to be available on the system:
- macOS:
pbcopy(built-in) - Linux/Wayland:
wl-copy - Linux/X11:
xcliporxsel - Windows:
clip.exe(built-in)
fetch --copy example.com/api/data
fetch --copy -o response.json example.com/api/dataDiscard the response body. Useful for checking status codes, viewing headers (with -v), or measuring timing (with --timing) without printing the response body.
Cannot be combined with --output, --remote-name, or --copy.
fetch --discard example.com
fetch --discard -v example.com # View headers only
fetch --discard --timing example.com # Measure timing onlyControl response formatting. Values: auto, on, off.
Markdown rendered on a terminal uses clickable OSC 8 hyperlinks when the terminal supports them. Terminal prose wraps to the detected terminal width, capped at 100 columns by default. Non-terminal output does not contain OSC 8 hyperlink escapes.
fetch --format off example.com # Disable formatting
fetch --format on example.com # Force formattingExtract a readable HTML/XHTML page as Markdown with YAML frontmatter. Markdown
responses (text/markdown and text/x-markdown) pass through after a url
frontmatter field. Article mode buffers at most 16 MiB of decoded content and
does not execute JavaScript. On terminals, embedded article images are fetched
and rendered unless --image off is set. Output files and pipes retain image
links in the raw Markdown. It cannot be combined with --discard,
--remote-name, or --remote-header-name.
The frontmatter fields are title, byline, site_name, published_time,
lang, dir, length, excerpt, and url when available. String values use
JSON quoting, which is also valid YAML. The final URL after redirects is used
for link resolution and the url field.
Output files and pipes receive raw, uncolored Markdown. --format, --color,
and --pager control only terminal presentation.
fetch --article https://example.com/story
fetch --article -o story.md https://example.com/storySelect content-encoding negotiation and streaming decoding: auto, br
(alias brotli), gzip, zstd, or off. auto advertises gzip, br, zstd;
the other enabled modes advertise only their selected encoding. off sends
no automatic encoding header and preserves encoded response bytes. An
explicit Accept-Encoding header is never replaced. The compatibility flag
--no-encode is equivalent to --compress off; selecting conflicting
explicit modes is an error.
Configure Encrypted ClientHello with auto, on, or off. The default is
off. auto discovers HTTPS/SVCB records with address resolution and uses a
validated advertised ECH configuration when one is available. Authenticated
DNS discovery failures are not silently downgraded. on requires a usable
advertised configuration, rejects explicit HTTP/3, and uses TCP with automatic
HTTP version selection. ECH requires TLS 1.3; explicit TLS 1.2 bounds are a
configuration error. WSS and TLS inspection use the same ECH discovery and
handshake reporting. ECH is not available through proxies or cleartext HTTP/2.
See Encrypted ClientHello for discovery and downgrade rules.
Record the final HTTP exchange in a HAR 1.2 sidecar at PATH. Standard output
(-) is not a valid HAR destination. The sidecar contains request headers,
cookies, credentials, and captured bodies. Protect it as sensitive data.
Request and response bodies are captured up to 16 MiB each. UTF-8 text is stored
as text and binary data as base64 while within the limit. For larger bodies,
the captured payload is omitted entirely and the HAR adds the comment Body omitted by fetch because it exceeds the 16 MiB HAR capture limit. HAR records
the final exchange after redirects and retries. It is not
available for WebSocket sessions, DNS/TLS inspection, gRPC discovery, or
--dry-run, and it cannot use the response output path.
Control colored output. Values: auto, on, off.
Alias: --colour
fetch --color off example.comControl image rendering. Values: auto, external, or off. auto uses
built-in decoders only. external allows bounded vips, magick, and
ffmpeg adapters after built-in decoding. The legacy native spelling
remains accepted as an alias for auto.
auto- Use only built-in decoders (JPEG, PNG, TIFF, WebP)external- Try built-in decoders, then bounded external adaptersoff- Disable image rendering
fetch --image auto example.com/image.png
fetch --image off example.com/photo.jpgChoose pager behavior with auto, on, or off. --no-pager remains a
compatibility alias for --pager off.
Accepted for compatibility. The Go implementation keeps deterministic alphabetical header rendering, so this option is currently a no-op.
Set WebSocket input handling to auto, text, or binary. This option is
valid only with a ws:// or wss:// URL. Text lines, interactive entries, and
incoming messages are bounded to 16 MiB; binary stdin is streamed in bounded
chunks. See WebSocket.
Print the portable Agent Skill without requiring a URL.
Manage the portable skill for auto, agents, codex, claude, gemini,
pi, or all. Use --scope user|project to select the destination and
--force to permit replacement/removal of modified files. --scope and
--force require an install or uninstall operation.
Disable piping output to a pager (less).
fetch --no-pager example.comUse a named session for persistent cookie storage across invocations. Cookies set by servers are saved to disk and automatically sent on subsequent requests using the same session name.
Session names must contain only alphanumeric characters, hyphens, and underscores ([a-zA-Z0-9_-]).
# First request — server sets cookies, they get saved
fetch --session api example.com/login -j '{"user":"me"}'
# Second request — saved cookies are sent automatically
fetch --session api example.com/dashboardSession files are stored in the user's cache directory:
- Linux:
~/.cache/fetch/sessions/<NAME>.json - macOS:
~/Library/Caches/fetch/sessions/<NAME>.json
Session files are limited to 2 MiB, and serialized cookie data is limited to 1 MiB. Each session can contain up to 2,048 cookies (64 per domain); cookie names and values are limited to 256 and 4,096 bytes. Cookies rejected or evicted by these limits produce a bounded warning.
Can also be configured per-host in the configuration file.
Timeout for the connection phase (DNS resolution, TCP connect, TLS handshake) in seconds. Accepts decimal values. Independent of --timeout, which covers the entire request.
fetch --connect-timeout 5 example.com
fetch --connect-timeout 5 --timeout 30 example.comRequest timeout in seconds. Accepts decimal values.
fetch --timeout 30 example.com
fetch --timeout 2.5 example.comMaximum automatic redirects. Default: 10. Use 0 to disable.
fetch --redirects 0 example.com # Don't follow redirects
fetch --redirects 10 example.comCredential-bearing headers, including custom headers with authentication-like names, are sent only within the initial origin. Header-name matching uses complete components rather than substrings. They are removed before a cross-origin redirected request and are not restored if a later redirect returns to the original origin.
When --cert configures mTLS, a redirect crossing a scheme, host, or port
boundary is refused before connecting to the destination. Same-origin
redirects remain allowed.
Request bodies are also protected at origin boundaries. A redirect is refused with an error before any bytes are sent to the destination when the effective redirected request has a body and the status/method semantics preserve that body. This rule is based on body presence, not a fixed list of methods, so it also covers body-bearing GET, HEAD, DELETE, OPTIONS, POST, and custom methods. There is no cross-origin body-replay opt-in. A 301 or 302 POST and every non-HEAD 303 are changed to GET without a body, so those redirects remain allowed.
Maximum number of retries for transient failures. Default: 0 (no retries).
The value must be between 0 and 100.
Retries occur on connection errors and retryable status codes (429, 502, 503, 504) for GET, HEAD, OPTIONS, and TRACE requests. Non-retryable errors (4xx, TLS certificate errors) are not retried. PUT and DELETE are not retried by default: although HTTP describes them as idempotent, individual APIs may implement side effects that are unsafe to repeat. POST, PATCH, PUT, DELETE, and custom methods require the explicit --retry-unsafe opt-in. Uses exponential backoff with jitter between attempts.
All attempts, redirects, response reads, bounded drains, and retry delays share one --timeout wall-clock budget. A request body must be replayable before a retry starts. Only the final attempt's response body is written to stdout. Retry notifications are printed to stderr (suppressed with --silent).
fetch --retry 3 example.com
fetch --retry 2 --retry-delay 0.5 example.comOpt in to automatic retries for methods other than GET, HEAD, OPTIONS, and
TRACE. This includes POST, PATCH, PUT, DELETE, and custom methods. Use it only
when the endpoint is known to tolerate replay, or when the request carries an
application-level idempotency guarantee such as a validated Idempotency-Key.
The option does not add an idempotency key or otherwise make a request safe.
fetch --method POST --data '{"action":"create"}' --retry 2 --retry-unsafe example.com/actionsDigest authentication is a separate challenge-response mechanism. A Digest
401 challenge can replay the request once (and a stale nonce can cause one
additional bounded replay) when --digest is enabled, including for an unsafe
method; --retry-unsafe does not control those authentication replays.
Initial delay between retries in seconds. Default: 1. Accepts decimal values.
The actual delay uses exponential backoff (delay doubles each attempt, capped at 30s) with ±25% jitter. If the server sends a Retry-After header, that value is used when it exceeds the computed delay and is capped at 30 seconds. A warning is printed when a server value is clamped. The request stops without sleeping when the remaining timeout cannot accommodate the delay.
fetch --retry 3 --retry-delay 2 example.com
fetch --retry 3 --retry-delay 0.5 example.comUse a custom DNS resolver. Supported endpoints are bare UDP addresses, or
udp://, tcp://, tls:///dot://, quic:///doq://, and https://
(DNS-over-HTTPS) endpoints. UDP and TCP default to port 53. DoT and DoQ
default to 853. DoH uses the URL port and requires HTTPS.
fetch --dns-server 8.8.8.8 example.com
fetch --dns-server tcp://1.1.1.1:53 example.com
fetch --dns-server dot://dns.example example.com
fetch --dns-server https://1.1.1.1/dns-query example.comEndpoints reject userinfo, fragments, invalid ports, and paths or queries on
non-DoH transports. Resolver endpoint parsing occurs during CLI/config
validation, before any network request. The configured resolver is shared by
HTTP/1.1, HTTP/2, forced HTTP/3, gRPC, WebSocket, TLS inspection, and update
traffic. Address candidates use Happy Eyeballs while preserving resolver
family preference. TCP and DoT queries use one operation-scoped pipelined
connection. DoQ uses one verified QUIC connection per resolver operation and
one bidirectional stream per DNS query. Hostname endpoints use a nonrecursive
platform bootstrap and negotiate the standard doq ALPN.
Inspect DNS resolution for the URL hostname only (no HTTP request is made). Without --dns-server, it queries the nameservers listed in the system resolver configuration (/etc/resolv.conf) directly, including on macOS. It reports every record type (A, AAAA, CNAME, TXT, MX, NS, SOA, SRV, CAA, SVCB, and HTTPS) with per-record TTLs, but does not apply macOS scoped, per-interface, VPN, or /etc/resolver routing. On platforms without a usable resolver file (notably Windows), or when the name is resolved only through OS mechanisms (the hosts file, NSS modules, or mDNS), it uses the platform resolver for A and AAAA records without per-record TTLs. If direct DNS returns no address records, platform-resolver addresses are added while any records already returned by direct DNS remain visible. With an explicit resolver it queries the same record types concurrently. The output includes resolver security, record counts, and duration. If one query fails, successful records remain visible, a warning identifies the incomplete record types, and the command exits with status 1.
fetch --inspect-dns example.com
fetch --inspect-dns --dns-server https://1.1.1.1/dns-query example.comRoute request through a proxy.
fetch --proxy http://localhost:8080 example.com
fetch --proxy socks5://localhost:1080 example.comConnect to the supplied IP address for a matching host and port while keeping
the URL host as the HTTP Host header and TLS SNI. Repeat the option or provide
comma-separated addresses to provide multiple candidates. A * host acts as
a fallback for any host. A leading + is accepted for curl compatibility.
fetch --resolve example.com:443:127.0.0.1 https://example.com
fetch --resolve '*:443:192.0.2.10' https://example.comMake request over a Unix domain socket. Unix-like systems only.
fetch --unix /var/run/docker.sock http://unix/containers/jsonMinimum TLS version. This is an alias for --min-tls. Values: 1.2, 1.3.
fetch --tls 1.3 example.comMinimum TLS version. Values: 1.2, 1.3.
fetch --min-tls 1.2 example.comMaximum TLS version. Values: 1.2, 1.3. Combine with --min-tls to allow a bounded range or require an exact TLS version.
fetch --min-tls 1.2 --max-tls 1.2 example.comInspect the TLS certificate chain by performing a TLS handshake only (no HTTP request is made). Displays the TLS version, cipher suite, ALPN protocol, full certificate chain with expiry status, Subject Alternative Names (SANs), and OCSP staple status. Requires an HTTPS URL. With --http 3, inspection uses a QUIC handshake and offers h3 ALPN. When ECH is enabled, inspection also reports the real or GREASE offer, acceptance or rejection, fallback status, and outer SNI. HTTP-only flags (e.g. --data, --timing, --grpc) are ignored with a warning.
fetch --inspect-tls example.com
fetch --inspect-tls --http 3 example.com
fetch --inspect-tls --insecure expired.badssl.comAccept invalid TLS certificates. Use with caution.
fetch --insecure https://self-signed.example.comCustom CA certificate file.
fetch --ca-cert ca-cert.pem example.comForce specific HTTP version. Values: 1, 2, 3.
1- HTTP/1.12- Force HTTP/2; direct HTTPS may otherwise try automatic HTTP/3 first3- HTTP/3 (QUIC)
When --http 2 is used with an http:// URL for gRPC requests, fetch automatically uses h2c (HTTP/2 over cleartext) to connect without TLS.
fetch --http 1 example.com
fetch --http 3 example.com
fetch --grpc --http 2 http://localhost:50051/pkg.Svc/Method # uses h2cThe compatibility aliases --http1, --http2, and --http3 select the
corresponding version without a value argument.
Disable automatic content-encoding negotiation and decoding. This is
equivalent to --compress off.
fetch --no-encode example.comRequest specific byte ranges. Can be used multiple times.
fetch -r 0-1023 example.com/file.bin
fetch -r 0-499 -r 1000-1499 example.com/file.binIncrease output verbosity. Can be stacked.
-v- Show response headers-vv- Show request and response headers with>/<prefixes-vvv- Show DNS and TLS details with>/</*prefixes
Diagnostic URLs and headers redact userinfo, sensitive query values, standard
credential headers, and credential-like custom header names. Query-key matching
is case-insensitive and covers names containing key, token, secret,
password, credential, signature, authorization, or session; names are
retained.
fetch -v example.com
fetch -vvv example.comDisplay a timing waterfall chart after the response. Shows DNS, TCP, TLS, TTFB, and body download phases as a proportional bar chart. Works independently of verbosity. Phases that don't apply (e.g., TLS for HTTP, TCP for HTTP/3, DNS/TCP/TLS for reused connections) are omitted.
fetch --timing https://example.com
fetch -T https://example.com
fetch --timing -vvv https://example.com # Both debug text and waterfallSuppress verbose output. Only errors shown on stderr.
fetch -s example.comDon't use HTTP status code for exit code. Always exit 0 on successful request.
fetch --ignore-status example.com/not-foundUse ws:// or wss:// URL schemes to open a WebSocket connection:
fetch ws://echo.websocket.events
fetch wss://echo.websocket.events -d "hello"Use --ws-interactive auto|on|off to control the terminal prompt.
Use --ws-message-mode auto|text|binary to select message types. See
WebSocket documentation for message limits, close handling,
proxy/TLS behavior, and unsupported options.
Enable gRPC mode. Automatically sets HTTP/2, POST method, and gRPC headers. When no local proto schema is provided, fetch automatically tries gRPC reflection before falling back to generic protobuf handling.
fetch --grpc https://localhost:50051/package.Service/MethodList available gRPC services. Uses reflection when a URL is provided, or runs offline when --proto-file / --proto-desc is provided.
fetch --grpc-list https://localhost:50051
fetch --grpc-list --proto-desc service.pbDescribe a gRPC service, method, or message. Accepts package.Service, package.Service/Method, package.Service.Method, and full message names.
fetch --grpc-describe grpc.health.v1.Health https://localhost:50051
fetch --grpc-describe grpc.health.v1.Health/Check --proto-desc service.pbCompile .proto file(s) for gRPC requests or offline discovery. Requires protoc. Can specify multiple comma-separated paths.
fetch --grpc --proto-file service.proto -j '{"field": "value"}' localhost:50051/pkg.Svc/MethodUse a pre-compiled descriptor set file instead of --proto-file.
# Generate descriptor:
protoc --descriptor_set_out=service.pb --include_imports service.proto
# Use descriptor:
fetch --grpc --proto-desc service.pb -j '{"field": "value"}' localhost:50051/pkg.Svc/MethodAdd import paths for proto compilation. Use with --proto-file.
fetch --grpc --proto-file service.proto --proto-import ./proto localhost:50051/pkg.Svc/MethodPlaintext servers are supported via h2c (HTTP/2 over cleartext) when using an http:// URL with HTTP/2. This works for --grpc and reflection-based discovery (--grpc-list, --grpc-describe).
Specify configuration file path.
fetch --config ~/.config/fetch/custom.conf example.comExecute a curl command using fetch. Parses a curl command string and translates its flags into the equivalent fetch options. The curl prefix is optional.
Cannot be combined with other request-specifying flags (URL, --method, --header, --data, auth flags, etc.). Meta flags like --dry-run, --color, --format, --no-pager, and --timing can still be used.
# Basic GET
fetch --from-curl 'curl https://example.com'
# POST with JSON
fetch --from-curl 'curl -X POST -H "Content-Type: application/json" -d {"key":"value"} https://example.com'
# With authentication
fetch --from-curl 'curl -u user:pass https://example.com'
# Follow redirects with retry
fetch --from-curl 'curl -L --max-redirs 5 --retry 3 https://example.com'
# Preview without sending
fetch --dry-run --from-curl 'curl -X PUT -d @data.json https://example.com'
# Without the curl prefix
fetch --from-curl 'https://example.com'Supported curl flags:
| Category | Curl Flags |
|---|---|
| Request | -X, -H, -d, --data-raw, --data-binary, --data-urlencode, --json, -F, -T, -I, -G |
| Auth | -u, --digest, --aws-sigv4, --oauth2-bearer |
| TLS | -k, --cacert, -E/--cert, --key, --tlsv1.x, --tls-max, `--ech hard |
| Output | -o, -O, -J |
| Network | -L, --max-redirs, -m/--max-time, --connect-timeout, -x, --unix-socket, --doh-url, --resolve, --retry, --retry-delay, --retry-unsafe, -r |
| HTTP version | -0, --http1.1, --http2, --http3 |
| Headers | -A, -e, -b |
| Verbosity | -v, -s |
| Protocol | --proto (restricts allowed protocols; errors if URL scheme is not allowed) |
| Default-compatible no-ops | --compressed, -S/--show-error, --fail-with-body, --no-keepalive |
| Presentation compatibility | -#/--progress-bar, --no-progress-meter |
Notes:
-b/--cookieonly supports inline cookie strings (e.g.,-b 'name=value'). Cookie jar files are not supported and will return an error.- A single
-d @filenameor-d @-body streams through fetch's native request-body path. Composite data bodies and--data-urlencode @filenameare materialized for curl compatibility and capped at 16 MiB. --data-urlencodesupports@filenameandname@filenameforms for reading and URL-encoding file contents. File bytes are encoded before any text conversion.-n/--netrcis not supported. Use--basic,--bearer, or an explicitAuthorizationheader instead.- Semantic curl flags that fetch cannot faithfully translate, such as
-f/--fail,-N/--no-buffer,--proto-default, and--proto-redir, return an error instead of being ignored.
Unknown curl flags return an error.
Print concise help information. Use -v --help for the full Markdown
reference. The detailed form uses the configured pager when appropriate.
--pager on forces paging, --pager off disables it, and NO_PAGER disables
automatic paging.
Print version.
Print build information as JSON. The output includes the fetch version, Go
version and build settings. Add
-v to include module dependency versions. Metadata commands read
configuration on a best-effort basis; an invalid config does not prevent help,
version, or build information output.
Check the latest GitHub release without downloading or replacing the executable. The command uses bounded HTTPS requests and carries only operational proxy, DNS, CA, and timeout settings.
Automatic checks run only for a validated normal request. They use the platform user cache interval and start a detached, silent updater child. The child does not inherit interactive standard input or terminal signals. Automatic failures are non-fatal to the request. Metadata-only commands, skill management, dry-runs, and updater children do not start another automatic check.
Update fetch binary in place. The selected archive must have a matching SHA-256
sidecar. Metadata is limited to 1 MiB, checksums to 1 KiB, and archives to 128
MiB. The archive is streamed to a temporary file and verified before extraction.
Use with --dry-run to validate release metadata, asset selection, checksum
availability, and executable preflight. Dry-run downloads bounded release
metadata and the checksum sidecar, but does not download the executable archive
or replace the binary. It does not update the automatic-check timestamp.
Output shell completion scripts. Values: bash, fish, powershell, zsh.
The scripts and dynamic candidates are generated from the same option registry
as CLI help. See Shell completion.
echo 'eval "$(fetch --complete bash)"' >> ~/.bashrc
fetch --complete zsh > ~/.zshrc.d/_fetch
fetch --complete fish > ~/.config/fish/completions/fetch.fishPrint request information without sending. The normalized absolute URL, method,
redacted headers, and at most 1,024 body bytes are shown. A preview never
consumes stdin or changes sessions, output files, HAR files, updater metadata,
or skill installations. When used with --update, checks release metadata and
destination preflight without downloading or installing.
fetch --dry-run -m POST -j '{"test": true}' example.com
fetch --update --dry-run--articlecannot be combined with WebSockets, gRPC, DNS/TLS inspection,--discard,--remote-name, or--remote-header-name.--harcannot be used with WebSockets, DNS/TLS inspection, gRPC discovery, dry-run, standard output, or the response output path. It supports unary gRPC only.- WebSockets require HTTP/1.1. They reject HTTP/2 and HTTP/3 forcing, Digest,
retries, retry delays, redirects, output, remote names, clipboard, clobber,
discard, range, compression controls (
--compressand--no-encode),--ignore-status, article, gRPC, and gRPC discovery options before connecting. - Explicit HTTP/3 cannot use a proxy or Unix socket. Forced HTTP/3 never falls back to TCP.
--ech oncannot use explicit HTTP/3 or TLS 1.2 bounds. ECH is not used through proxies or cleartext HTTP/2.--sort-headersis accepted as a compatibility no-op. Go's normal HTTP stack uses deterministic header serialization and does not preserve input wire order.
See Limits and safety for all body, protocol, archive, and subprocess caps.
| Variable | Description |
|---|---|
AWS_ACCESS_KEY_ID |
AWS access key for --aws-sigv4 |
AWS_SECRET_ACCESS_KEY |
AWS secret key for --aws-sigv4 |
AWS_SESSION_TOKEN |
Optional temporary AWS session token for --aws-sigv4 |
VISUAL / EDITOR |
Editor for --edit option |
HTTPS_PROXY / https_proxy |
HTTPS/WSS proxy URL |
HTTP_PROXY / http_proxy |
HTTP/WS proxy URL |
ALL_PROXY / all_proxy |
Fallback proxy URL |
NO_PROXY / no_proxy |
Hosts, domains, IPs, CIDR ranges, or ports |
Many options support file references with the @ prefix:
@filename- Read content from file@-- Read content from stdin@~/path- Home directory expansion
fetch -j @data.json -m POST example.com
echo '{"test": true}' | fetch -j @- -m POST example.com