Skip to content
Switch branches/tags
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Chat on Freenode PyPI version License Documentation Status

A command-line utility that runs code blocks from documentation written in markdown.


This utility allows you to run your markdown files as if they were scripts. Every code snippet with code highlighting tag is run with interpreter defined by the tag.

Why do this?

We had a very long installation documentation for our project and needed a quick way of testing it. That's how rundoc is born. It is now a general purpose tool that can be used for multiple purposes like executing a tutorial documentation, using docs as a script, etc.


install from pypi (recommend)

pip3 install rundoc

install from git (latest master)

pip3 install -U git+


Rundoc collects fenced code blocks from input markdown file and executes them in same order as they appear in the file. Interpreter is specified by the highlight tag of the code block.

Example of fenced code block in markdown file:

for x in `seq 0 10`; do
    echo $x
    sleep 1

Interpreter will be automatically selected using the highlight tag of the code block (in above example bash). Code is piped to the interpreter through stdin. If highlight tag is not specified, rundoc will ignore that code block.

Rundoc can save json file after execution, which contains all code blocks and their outputs. You can also replay all the actions by running this output file.

WARNING: Each code block is treated as separate step and is run in fresh interpreter session! Variables that you exported in the code block will not be passed to next code block. You can specify global variables in env code block (see Environment variables below).

Run markdown file

Execute code blocks in file:

rundoc run

Rundoc will execute all fenced code blocks that contain highlight tag.

Get a prompt

  • You will be prompted before executing each code block with ability to modify the input.
  • When done reviewing/modifying the code block, press Return to execute it and move to the next one.
  • Program will exit when last code block is finished executing or when you press ctrl+c.

Get prompts

Since version 0.3.0, rundoc will just run everything by default, prompting you only for missing variables at the beginning. you can use single or multiple -a options to enable more prompts:

  • -a - Ask (prompt) for all variables.
  • -aa - Also ask to modify code block on failure and start over at that step.
  • -aaa - Also prompt with each code block before executing it.
rundoc run -aaa

The above will show each code block and allow you to modify it and press <RETURN> to run it.


You can force the prompt on specific steps using -b or --breakpoint option followed by a step number. You can use this option multiple times to add multiple breakpoints. When used, rundoc will stop at the selected steps and prompt user to modify code regardless of the --ask setting.


If you need to add a delay between codeblocks, you can add -p or --pause option to specify number of seconds for the puase. (This option does nothing in conjunction with -aaa):

rundoc run -p 2

Some step fails first couple of times but it's normal? That may happen and you would just want to retry that step a couple of times. To do so use -r or --retry option followed by max number of retries and rundoc will run the same step again until it succeeds or reaches max retries in which case it will exit:

rundoc run -r 10

But you don't want it to retry right away, correct? You can specify a delay between each try with -P (capital P) or --retry-pause option followed by number of seconds:

rundoc run -r 10 -P 2

Start from specific step

You can start at specified step using -s or --step option:

rundoc run -s5

This is useful when your N-th code block fails and rundoc exits and you want to continue from that step.

Save output

Output can be saved as a json file containing these fields:

  • env (dict): dictionary of set environment variables for the session
  • code_blocks (list): list of code blocks that contains the following
    • code (str): original code
    • interpreter (str): interpreter used for this code block
    • tags (str): tags as they appear in markdown file
    • runs (list): list of run attempts
      • output (str): merged stdout and stderr of the code block execution
      • retcode (int): exit code of the code block
      • time_start (float): timestamp when execution started (seconds from epoch)
      • time_stop (float): timestamp when execution finished (seconds from epoch)
      • user_code (str): code that user actually executed with prompt

To save output use -o or --output option when running rundoc:

rundoc run -o output.json


By default, rundoc executes all fenced code blocks that have highlithing tag set in markdown file. If you want to limit execution to subset of the code blocks, use tags. Tags can be specified with -t or --tags option followed by hash (#) separated list of tags:

rundoc run -t bash#python3

This will execute only those code blocks that have at least one of the specified highlight tags: in this example only bash and python code blocks.

If you want to further isolate code blocks of the same highlight tag, you can use rundoc tag syntax, e.g.:

echo "custom-tagged code block"

In this syntax, multiple tags are applied to same code block and are separated with hash symbol #. In the example above there are 4 tags: bash, custom-branch, v2 and test. First tag always defines the interpreter. If any of it's tags is specified by -t or --tags option, it will be executed. Code blocks that do not contain at least one of the specified tags will be skipped.

More tags

In addition to -t or --tags option, you can also use the following 2 options to furthere fine-tune your code block filtering:

  • -T or --must-have-tags - same as --tags but it requires all listed tags to be present in the markdown code block or it will be skipped. The order of tags is not important.
  • -N, or --must-not-have-tags - same as --tags but it requres that none of the listed tags is present in the markdown code block. It is used to filter out unwanted ones.

You can use any of the tags features individually or combine them.

Environment variables

Define required environment variables anywhere in the documentaion with a special code block tagged as env or environment at the beginning:

  • As in example above, define variables one on each line.
  • When you run the docs with -a option, you will be prompted for all of those.
  • Empty values (e.g. var1 and var2 in example) will try to collect actual values from your system environment, so if var1 was exported before you ran the docs, it will collect it's value as the default value and will not prompt you for it if -a option was not used.
  • All variables will be passed to env for every code block that's being executed.
  • If you use rundoc with tag option -t, environment blocks will be filtered in the same way as code blocks.
  • Code blocks will only be selected if no tags are specified on execution or any of the selected code blocks contains at least one of the env's tags (in the above case version5).


You can define required credentials or other secrets anywhere in the documentaion as a special code block tagged as secret or secrets at the beginning:


Secrets behave just as env blocks with one single difference: they are never saved in the output file and are expected to be empty in markdown file so that user must provide them during execution. If you want to use rundoc as part of automation and can't input secrets by hand, you can always export them beforehand and use -i option (see next section).

Action tags

Action tags are special tags that are used as a first tag instead of interpreter. Code blocks with these tags are not going to be executed, but will be used to perform specific actions, like creating a file.

To get a list of available action tags and their usage in current version run: rundoc action-tags. New action tags will be included often in new versions of rundoc.

Force variable collection

You can force rundoc to check if any of the variables defined with env tag is already exported in your current system environment and use it's value instead of the one defined in markdown file. To do this use -i or --inherit-env option when running rundoc. The list of variables that is presented to you when you run rundoc and that will be used in the session will now contain values defined in the system environment.

export var2=system_value_2
export var3=system_value_3
rundoc run -i

Single session

By default, all code blocks are run in separate interpreter sessions. If you define a function or set a variable in one code block, it will not be available in next one.
To work around this use --single-session option followed by the name of the interpreter. All code blocks will be merged into a single step.
Note that only one interpreter type can be run per markdown file if you use single session option.

rundoc run --single-session bash


To replay all code blocks found in output of run command, just use replay command like so:

rundoc replay output.json

The above command will just turn last runs of each code block into a new code block and run them. It will ignore all run tries that did not succeed at first. Last run may have original command or user modified one and replay does not care about that, it just runs the last command it finds in each code block.

You can still use -aaa, -p, -s, -o, -r, and -P options with replay command:

rundoc replay -s 2 -p 1 -r 20 -P 5 output.json -o replay_output.json

Tips and tricks

Color output

Are you using light terminal background and can't see sh*t? Use rundoc with --light option and save your eyesight!

List tags

You can list all unique tags that appear in the file and their counts by using list-tags command:

rundoc list-tags

List code blocks

Wouldn't it be great to be able to list all code blocks that are going to be executed before actually using run command? You can! To print json file similar to output but without actually running anything you can use list-blocks command:

rundoc list-blocks -t bash -T tag1#tag2#tag3 -N tag4#tag5

or add --pretty option to have human readable output:

rundoc list-blocks -t bash -T tag1#tag2#tag3 -N tag4#tag5 --pretty

Similar projects

List of similar projects that I found:

If you bump into more, let me know.


A command-line utility that runs code blocks from documentation.




No packages published