1. hub-api(1)
  2. hub manual
  3. hub-api(1)

Low-level GitHub API request interface.


hub api [-it] [-X METHOD] [-H HEADER] [--cache TTL] ENDPOINT [-F FIELD|--input FILE]


-X, --method METHOD

The HTTP method to use for the request (default: "GET"). The method is automatically set to "POST" if --field, --raw-field, or --input are used.

Use -XGET to force serializing fields into the query string for the GET request instead of JSON body of the POST request.

-F, --field KEY=VALUE

Data to serialize with the request. VALUE has some magic handling; use --raw-field for sending arbitrary string values.

If VALUE starts with "@", the rest of the value is interpreted as a filename to read the value from. Use "@-" to read from standard input.

If VALUE is "true", "false", "null", or looks like a number, an appropriate JSON type is used instead of a string.

It is not possible to serialize VALUE as a nested JSON array or hash. Instead, construct the request payload externally and pass it via --input.

Unless -XGET was used, all fields are sent serialized as JSON within the request body. When ENDPOINT is "graphql", all fields other than "query" are grouped under "variables". See https://graphql.org/learn/queries/#variables

-f, --raw-field KEY=VALUE

Same as --field, except that it allows values starting with "@", literal strings "true", "false", and "null", as well as strings that look like numbers.

--input FILE

The filename to read the raw request body from. Use "-" to read from standard input. Use this when you want to manually construct the request payload.

-H, --header KEY:VALUE

Set an HTTP request header.

-i, --include

Include HTTP response headers in the output.

-t, --flat

Parse response JSON and output the data in a line-based key-value format suitable for use in shell scripts.


Automatically request and output the next page of results until all resources have been listed. For GET requests, this follows the <next\> resource as indicated in the "Link" response header. For GraphQL queries, this utilizes pageInfo that must be present in the query; see EXAMPLES.

Note that multiple JSON documents will be output as a result. If the API rate limit has been reached, the final document that is output will be the HTTP 403 notice, and the process will exit with a non-zero status. One way this can be avoided is by enabling --obey-ratelimit.


Enable colored output even if stdout is not a terminal. WHEN can be one of "always" (default for --color), "never", or "auto" (default).

--cache TTL

Cache valid responses to GET requests for TTL seconds.

When using "graphql" as ENDPOINT, caching will apply to responses to POST requests as well. Just make sure to not use --cache for any GraphQL mutations.


After exceeding the API rate limit, pause the process until the reset time of the current rate limit window and retry the request. Note that this may cause the process to hang for a long time (maximum of 1 hour).


The GitHub API endpoint to send the HTTP request to (default: "/").

To learn about available endpoints, see https://developer.github.com/v3/. To make GraphQL queries, use "graphql" as ENDPOINT and pass -F query=QUERY.

If the literal strings "{owner}" or "{repo}" appear in ENDPOINT or in the GraphQL "query" field, fill in those placeholders with values read from the git remote configuration of the current git repository.


# fetch information about the currently authenticated user as JSON
$ hub api user

# list user repositories as line-based output
$ hub api --flat users/octocat/repos

# post a comment to issue #23 of the current repository
$ hub api repos/{owner}/{repo}/issues/23/comments --raw-field 'body=Nice job!'

# perform a GraphQL query read from a file
$ hub api graphql -F query=@path/to/myquery.graphql

# perform pagination with GraphQL
$ hub api --paginate graphql -f query='
  query($endCursor: String) {
    repositoryOwner(login: "USER") {
      repositories(first: 100, after: $endCursor) {
        nodes {
        pageInfo {

See also