11. Command Environment Variables

When you send a command invocation to Cog, it can have many arguments and options. There are also several other pieces of important metadata that can be useful for commands to be aware of during execution. To get these data into the actual executable running that command, we expose them as environment variables. While libraries like cog-rb and pycog3 provide language-friendly abstractions to interact with these data, it can be useful to know the details of how they are structured, in case you need to perform some unique processing, or are writing commands in a language where a Cog library does not yet exist.

In general, all Cog-related environment variables will be prefixed with COG_.

11.1. Arguments

Arguments to Cog commands are represent in much the same way as they are in traditional UNIX shell scripting.

  • COG_ARGC: the number of arguments
  • COG_ARGV_<N>: The nth argument, starting at N = 0

For example, given the notional command invocation my:command foo bar baz, the executable would access the arguments through the following environment variables:


11.2. Options

As they behave like a map, options are handled a bit differently

  • COG_OPTS: a comma-delimited list of all option names
  • COG_OPT_<NAME>: the value of the NAME option

For example: the executable processing the invocation my:command --verbose --force --foo=bar would access the options through the following variables:


Note that the names given in COG_OPTS are lowercase, while the names of all environment variables provided by Cog are all uppercase.

11.2.1. List-valued Options

The above model works fine for options that have scalar values. For option values that are lists, we use an approach that is a hybrid between the argument processing and the scalar option processing we discussed above.

  • COG_OPT_<NAME>_COUNT: The number of values for the NAME option
  • COG_OPT_<NAME>_<N>: The nth value of the NAME option, starting at N = 0.

For an invocation like my:command --foo=one --foo=two --foo=three, the executable would access the values of foo with these variables:



Only scalar-valued option values will have a COG_OPT_<NAME> environment variable. Variables pertaining to list-valued options will always end in either _COUNT or _N (where N is an integer).

11.3. General Metadata

The following variables are set for all commands.

  • COG_BUNDLE: The name of the bundle the current command is a member of.
  • COG_CHAT_HANDLE: the chat handle of the user executing the command. This is the bare handle; if I am @user in Slack, the value will be just "user"
  • COG_COMMAND: The name of the current command being executed. Does not include the bundle; that is, when executing the command twitter:tweet, for example, COG_COMMAND will equal "tweet".
  • COG_INVOCATION_ID: a unique ID string identifying the specific invocation within a pipeline stage .
  • COG_INVOCATION_STEP: the string "first" if this is the first invocation of a command for a given stage of a pipeline, the string "last" if it is the last invocation, and unset if any intermediate invocation. If there is only a single invocation to be run in a given stage, the value will be "last".
  • COG_PIPELINE_ID: a unique ID string assigned to each pipeline. All commands in a given pipeline will receive the same value.
  • COG_ROOM: the chat room where the command was invoked. This is the bare room name; if it was executed in the #general channel in Slack, the value would be just "general". For 1-on-1 interactions, this will be the string "direct".
  • COG_SERVICE_TOKEN: a unique token for accessing Cog’s service infrastructure.
  • COG_SERVICES_ROOT: Host and port for accessing Cog’s service infrastructure.

All other environment variables (e.g., PATH, USER, etc.) are inherited from the Relay process.