SSH commands as API microservices

This is a quick guide on how to turn SSH commands into a REST API service. The use-case may be remote administration of devices or equipment that does not offer a REST interface or making sure that access to SSH commands is restricted to selected external REST-based API clients only.


The first thing needed is code of the service that will connect to SSH servers. Below is a service doing just that - it receives name of the command to execute and host to run in on, translating stdout and stderr of SSH commands into response documents which Zato in turn serializes to JSON.

# -*- coding: utf-8 -*-

# stdlib
from traceback import format_exc

# Zato
from zato.server.service import Service

class SSHInvoker(Service):
    """ Accepts an SSH command to run on a remote host and returns its output to caller.

    # A list of elements that we expect on input
    input = 'host', 'command'

    # A list of elements that our responses will contain
    output = 'is_ok', 'cid', '-stdout', '-stderr'

    def handle(self):

        # Local aliases
        host =
        command = self.request.input.command

        # Correlation ID is always returned
        self.response.payload.cid = self.cid

            # Build the full command
            full_command = f'ssh {host} {command}'

            # Run the command and collect output
            output = self.commands.invoke(full_command)

            # Assign both stdout and stderr to response
            self.response.payload.stdout = output.stdout
            self.response.payload.stderr = output.stderr

        except Exception:
            # Catch any exception and log it
            self.logger.warn('Exception caught (%s), e:`%s', self.cid, format_exc())

            # Indicate an error
            self.response.payload.is_ok = False

            # Everything went fine
            self.response.payload.is_ok = True


In the Zato Dashboard, let's go ahead and create an HTTP Basic Auth definition that a remote API client will authenticate against:

Now, the SSH service can be mounted on a newly created REST channel - note the security definition used and that data format is set to JSON. We can skip all the other details such as caching or rate limiting, for illustration purposes, this is not needed.


At this point, everything is ready to use. We could make it accessible to external API clients but, for testing purposes, let's simply invoke our SSH API gateway service from the command line:

$ curl "api:password@localhost:11223/api/ssh" -d \
    '{"host":"localhost", "command":"uptime"}'
    "is_ok": true,
    "cid": "27406f29c66c2ab6296bc0c0",
    "stdout": " 09:45:42 up 37 min,  1 user,  load average: 0.14, 0.27, 0.18\n"}
Note that, at this stage, the service should be used in trusted environments only, e.g. it will run any command that it is given on input which means that in the next iteration it could be changed to only allow commands from an allow-list, rejecting anything that is not recognized.

And this completes it - the service is deployed and made accessible via a REST channel that can be invoked using JSON. Any command can be sent to any host and their output will be returned to API callers in JSON responses.

Next steps