diff --git a/.travis.yml b/.travis.yml
index f7f4401..2a5ec73 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -29,6 +29,8 @@ git:
 #script:
 #  - julia -e 'Pkg.clone(pwd()); Pkg.build("Examples"); Pkg.test("Examples"; coverage=true)'
 after_success:
+  # build docs
+  - julia -e 'cd(Pkg.dir("Examples")); Pkg.add("Documenter"); include("docs/make.jl")'
   # push coverage results to Coveralls
   - julia -e 'cd(Pkg.dir("Examples")); Pkg.add("Coverage"); using Coverage; Coveralls.submit(Coveralls.process_folder())'
   # push coverage results to Codecov
diff --git a/REQUIRE b/REQUIRE
index 289234b..d4895a9 100644
--- a/REQUIRE
+++ b/REQUIRE
@@ -1,2 +1,3 @@
 julia 0.6
 JSON
+Compat
diff --git a/docs/.gitignore b/docs/.gitignore
new file mode 100644
index 0000000..100775b
--- /dev/null
+++ b/docs/.gitignore
@@ -0,0 +1,3 @@
+build/
+site/
+generated/
diff --git a/docs/make.jl b/docs/make.jl
new file mode 100644
index 0000000..022a76b
--- /dev/null
+++ b/docs/make.jl
@@ -0,0 +1,32 @@
+using Revise
+using Documenter
+using Examples
+
+# generate examples
+EXAMPLE = joinpath(@__DIR__, "..", "examples", "example.jl")
+OUTPUT = joinpath(@__DIR__, "src/generated")
+
+Examples.markdown(EXAMPLE, OUTPUT)
+Examples.notebook(EXAMPLE, OUTPUT)
+Examples.script(EXAMPLE, OUTPUT)
+
+makedocs(
+    modules = [Examples],
+    format = :html,
+    sitename = "Examples.jl",
+    pages = Any[
+        "index.md",
+        "fileformat.md",
+        "pipeline.md",
+        "outputformats.md",
+        "customprocessing.md",
+        "documenter.md",
+        "generated/example.md"]
+)
+
+deploydocs(
+    repo = "github.com/fredrikekre/Examples.jl.git",
+    target = "build",
+    deps = nothing,
+    make = nothing
+)
diff --git a/docs/src/customprocessing.md b/docs/src/customprocessing.md
new file mode 100644
index 0000000..965b780
--- /dev/null
+++ b/docs/src/customprocessing.md
@@ -0,0 +1,47 @@
+# [**5.** Custom pre- and post-processing](@id Custom-pre-and-post-processing)
+
+Since all packages are different, and may have different demands on how
+to create a nice example for the documentation it is important that
+the package maintainer does not feel limited by the by default provided syntax
+that this package offers. While you can generally come a long way by utilizing
+[line filtering](@ref Filtering-lines) there might be situations where you need
+to manually hook into the generation and change things. In `Examples.jl` this
+is done by letting the user supply custom pre- and post-processing functions
+that may do transformation of the content.
+
+All of the generators ([`Examples.markdown`](@ref), [`Examples.notebook`](@ref)
+and [`Examples.script`](@ref)) accepts `preprocess` and `postprocess` keyword
+arguments. The default "transformation" is the `identity` function. The input
+to the transformation functions is a `String`, and the output should be the
+transformed `String`.
+
+`preprocess` is sent the raw input that is read from the source file ([modulo the
+default line ending transformation](@ref Pre-processing)). `postprocess` is given
+different things depending on the output: For markdown and script output `postprocess`
+is given the content `String` just before writing it to the output file, but for
+notebook output `postprocess` is given the dictionary representing the notebook,
+since, in general, this is more useful.
+
+As an example, lets say we want to splice the date of generation into the output.
+We could of course update our source file before generating the docs, but we could
+instead use a `preprocess` function that splices the date into the source for us.
+Consider the following source file:
+```julia
+#' # Example
+#' This example was generated DATEOFTODAY
+
+x = 1 // 3
+```
+where `DATEOFTODAY` is a placeholder, to make it easier for our `preprocess` function
+to find the location. Now, lets define the `preprocess` function, for example
+```julia
+function update_date(content)
+    content = replace(content, "DATEOFTODAY" => Date(now()))
+    return content
+end
+```
+which would replace every occurrence of `"DATEOFTODAY"` with the current date. We would
+now simply give this function to the generator, for example:
+```julia
+Examples.markdown("input.jl", "outputdir"; preprocess = update_date)
+```
diff --git a/docs/src/documenter.md b/docs/src/documenter.md
new file mode 100644
index 0000000..97356e6
--- /dev/null
+++ b/docs/src/documenter.md
@@ -0,0 +1,3 @@
+# [**6.** Interaction with Documenter.jl](@id Interaction-with-Documenter)
+
+TBW
diff --git a/docs/src/fileformat.md b/docs/src/fileformat.md
new file mode 100644
index 0000000..1d08253
--- /dev/null
+++ b/docs/src/fileformat.md
@@ -0,0 +1,111 @@
+# **2.** File Format
+
+The source file format for `Examples.jl` is a regular, commented, julia (`.jl`) scripts.
+The idea is that the scripts also serve as documentation on their own and it is also
+simple to include them in the test-suite, with e.g. `include`, to make sure the examples
+stay up do date with other changes in your package.
+
+## [**2.1.** Syntax](@id Syntax)
+
+The basic syntax is simple:
+- lines starting with `#'` is treated as markdown,
+- all other lines are treated as julia code.
+
+The reason for using `#'` instead of `#` is that we want to be able to use `#` as comments,
+just as in a regular script. Lets look at a simple example:
+```julia
+#' # Rational numbers
+#'
+#' In julia rational numbers can be constructed with the `//` operator.
+#' Lets define two rational numbers, `x` and `y`:
+
+x = 1//3
+y = 2//5
+
+#' When adding `x` and `y` together we obtain a new rational number:
+
+z = x + y
+```
+In the lines `#'` we can use regular markdown syntax, for example the `#`
+used for the heading and the backticks for formatting code. The other lines are regular
+julia code. We note a couple of things:
+- The script is valid julia, which means that we can `include` it and the example will run
+- The script is "self-explanatory", i.e. the markdown lines works as comments and
+  thus serve as good documentation on its own.
+
+For simple use this is all you need to know, the script above is valid. Let's take a look
+at what the above snippet would generate, with default settings:
+
+- [`Examples.markdown`](@ref): leading `#'` are removed, and code lines are wrapped in
+  `@example`-blocks:
+  ````markdown
+  # Rational numbers
+
+  In julia rational numbers can be constructed with the `//` operator.
+  Lets define two rational numbers, `x` and `y`:
+
+  ```@example filename
+  x = 1//3
+  y = 2//5
+  ```
+
+  When adding `x` and `y` together we obtain a new rational number:
+
+  ```@example filename
+  z = x + y
+  ```
+  ````
+
+- [`Examples.notebook`](@ref): leading `#'` are removed, markdown lines are placed in
+  `"markdown"` cells, and code lines in `"code"` cells:
+  ```
+           │ # Rational numbers
+           │
+           │ In julia rational numbers can be constructed with the `//` operator.
+           │ Lets define two rational numbers, `x` and `y`:
+
+  In [1]:  │ x = 1//3
+           │ y = 2//5
+
+  Out [1]: │ 2//5
+
+           │ When adding `x` and `y` together we obtain a new rational number:
+
+  In [2]:  │ z = x + y
+
+  Out [2]: │ 11//15
+  ```
+
+- [`Examples.script`](@ref): all lines starting with `#'` are removed:
+  ```julia
+  x = 1//3
+  y = 2//5
+
+  z = x + y
+  ```
+
+## [**2.2.** Filtering Lines](@id Filtering-lines)
+
+It is possible to filter out lines depending on the output format. For this purpose,
+there are three different "tokens" that can be placed on the start of the line:
+- `#md`: markdown output only,
+- `#nb`: notebook output only,
+- `#jl`: script output only.
+
+Lines starting with one of these tokens are filtered out in the
+[preprocessing step](@ref Pre-processing).
+
+Suppose, for example, that we want to include a docstring within a `@docs` block
+using Documenter. Obviously we don't want to include this in the notebook,
+since `@docs` is Documenter syntax that the notebook will not understand. This
+is a case where we can prepend `#md` to those lines:
+````julia
+#md #' ```@docs
+#md #' Examples.markdown
+#md #' Examples.notebook
+#md #' Examples.markdown
+#md #' ```
+````
+The lines in the example above would be filtered out in the preprocessing step, unless we are
+generating a markdown file. When generating a markdown file we would simple remove
+the leading `#md ` from the lines. Beware that the space after the tag is also removed.
diff --git a/docs/src/index.md b/docs/src/index.md
new file mode 100644
index 0000000..a8aa3a6
--- /dev/null
+++ b/docs/src/index.md
@@ -0,0 +1,51 @@
+# **1.** Introduction
+
+Welcome to the documentation for `Examples.jl`. A simplistic package
+to help you organize examples for you package documentation.
+
+### What?
+
+`Examples.jl` is a package that, based on a single source file, generates markdown,
+for e.g. [Documenter.jl](https://github.com/JuliaDocs/Documenter.jl),
+[Jupyter notebooks](http://jupyter.org/) and uncommented scripts for documentation
+of your package.
+
+The main design goal is simplicity. It should be simple to use, and the syntax should
+be simple. In short all you have to do is to write a commented julia script!
+
+The package consists mainly of three functions, which all takes the same script file
+as input, but generates different output:
+- [`Examples.markdown`](@ref): generates a markdown file
+- [`Examples.notebook`](@ref): generates an (optionally executed) notebook
+- [`Examples.script`](@ref): generates a plain script file, removing everything
+  that is not code
+
+### Why?
+
+Examples are (probably) the best way to showcase your awesome package, and examples
+are often the best way for a new user to learn how to use it. It is therefore important
+that the documentation of your package contains examples for users to read and study.
+However, people are different, and we all prefer different ways of trying out a new
+package. Some people wants to RTFM, others want to explore the package interactively in,
+for example, a notebook, and some people wants to study the source code. The aim of
+`Examples.jl` is to make it easy to give the user all of these options, while still
+keeping maintenance to a minimum.
+
+It is quite common that packages have "example notebooks" to showcase the package.
+Notebooks are great for this, but they are not so great with version control, like git.
+The reason is that a notebook is a very "rich" format since it contains output and other
+metadata. Changes to the notebook thus result in large diffs, which makes it harder to
+review the actual changes.
+
+It is also common that packages include examples in the documentation, for example
+by using [Documenter.jl](https://github.com/JuliaDocs/Documenter.jl) `@example`-blocks.
+This is also great, but it is not quite as interactive as a notebook, for the users
+who prefer that.
+
+`Examples.jl` tries to solve the problems above by creating the output as a part of the doc
+build. `Examples.jl` generates the output from a single source file which makes it easier to
+maintain, test, and keep the manual and your example notebooks in sync.
+
+### How?
+
+TBD
diff --git a/docs/src/outputformats.md b/docs/src/outputformats.md
new file mode 100644
index 0000000..836f19e
--- /dev/null
+++ b/docs/src/outputformats.md
@@ -0,0 +1,50 @@
+# [**4.** Output formats](@id Output-formats)
+
+
+## [**4.1.** Markdown output](@id Markdown-output)
+
+````
+#' # Markdown                                                  ┐
+#'                                                             │
+#' This line is treated as markdown, since it starts with #'   │
+#' The leading #' (including the space) is removed             ┘
+
+#' Here is an example with some code                           ]
+
+x = sin.(cos.([1, 2, 3]))                                      ┐
+y = x.^2 - x                                                   ┘
+````
+
+By default, `CodeChunks` written to Documenter `@example` blocks. For example,
+the code above would result in the following markdown:
+
+````markdown
+# Markdown
+
+This line is treated as markdown, since it starts with #'
+The leading #' (including the space) is removed
+
+Here is an example with some code
+
+```@example
+x = sin.(cos.([1, 2, 3]))
+y = x.^2 - x
+```
+````
+
+```@docs
+Examples.markdown
+```
+
+## [**4.2.** Notebook output](@id Notebook-output)
+
+```@docs
+Examples.notebook
+```
+
+
+## [**4.3.** Script output](@id Script-output)
+
+```@docs
+Examples.script
+```
diff --git a/docs/src/pipeline.md b/docs/src/pipeline.md
new file mode 100644
index 0000000..916bf5e
--- /dev/null
+++ b/docs/src/pipeline.md
@@ -0,0 +1,136 @@
+# **3.** Processing pipeline
+
+The generation of output follows the same pipeline for all output formats:
+1. [Pre-processing](@ref)
+2. [Parsing](@ref)
+3. [Document generation](@ref)
+4. [Post-processing](@ref)
+5. [Writing to file](@ref)
+
+
+## [**3.1.** Pre-processing](@id Pre-processing)
+
+The first step is pre-processing of the input file. The file is read to a `String`
+and CRLF style line endings (`"\r\n"`) are replaced with LF line endings (`"\n"`) to simplify
+internal processing. The next step is to apply the user specified pre-processing function.
+See [Custom pre- and post-processing](@ref Custom-pre-and-post-processing).
+
+Next the line filtering is performed, see [Filtering lines](@ref), meaning that lines
+starting with `#md `, `#nb ` or `#jl ` are handled (either just the token itself is removed,
+or the full line, depending on the output target).
+
+
+## [**3.2.** Parsing](@id Parsing)
+
+After the preprocessing the file is parsed. The first step is to categorize each line
+and mark them as either markdown or code according to the rules described in the
+[Syntax](@ref Syntax) section. Lets consider the example from the previous section
+with each line categorized:
+```
+#' # Rational numbers                                                     <- markdown
+#'                                                                        <- markdown
+#' In julia rational numbers can be constructed with the `//` operator.   <- markdown
+#' Lets define two rational numbers, `x` and `y`:                         <- markdown
+                                                                          <- code
+x = 1 // 3                                                                <- code
+y = 2 // 5                                                                <- code
+                                                                          <- code
+#' When adding `x` and `y` together we obtain a new rational number:      <- markdown
+                                                                          <- code
+z = x + y                                                                 <- code
+```
+
+In the next step the lines are grouped into "chunks" of markdown and code.
+This is done by simply collecting adjacent lines of the same "type" into
+chunks:
+```
+#' # Rational numbers                                                     ┐
+#'                                                                        │
+#' In julia rational numbers can be constructed with the `//` operator.   │ markdown
+#' Lets define two rational numbers, `x` and `y`:                         ┘
+                                                                          ┐
+x = 1 // 3                                                                │
+y = 2 // 5                                                                │ code
+                                                                          ┘
+#' When adding `x` and `y` together we obtain a new rational number:      ] markdown
+                                                                          ┐
+z = x + y                                                                 ┘ code
+```
+
+In the last parsing step all empty leading and trailing lines for each chunk
+are removed, but empty lines *within the same* block are kept. The leading `#' `
+tokens are also removed from the markdown chunks. Finally we would
+end up with the following 4 chunks:
+
+Chunks #1:
+```markdown
+# Rational numbers
+
+In julia rational numbers can be constructed with the `//` operator.
+Lets define two rational numbers, `x` and `y`:
+```
+Chunk #2:
+```julia
+x = 1 // 3
+y = 2 // 5
+```
+Chunk #3:
+```markdown
+When adding `x` and `y` together we obtain a new rational number:
+```
+Chunk #4:
+```julia
+z = x + y
+```
+
+It is then up to the [Document generation](@ref) step to decide how these chunks should be treated.
+
+### Custom control over chunk splits
+
+Sometimes it is convenient to be able to manually control how the chunks are split.
+For example, if you want to split a block of code into two, such that they end up in
+two different `@example` blocks or notebook cells. The `#-` token can be used for this
+purpose. All lines starting with `#-` are used as "chunk-splitters":
+```julia
+x = 1 // 3
+y = 2 // 5
+#-
+z = x + y
+```
+The example above would result in two consecutive code-chunks.
+
+!!! tip
+    The rest of the line, after `#-`, is discarded, so it is possible to use e.g.
+    `#-------------` as a chunk splitter, which may make the source code more readable.
+
+
+## [**3.3.** Document generation](@id Document-generation)
+
+After the parsing it is time to generate the output. What is done in this step is
+very different depending on the output target, and it is describe in more detail in
+the Output format sections: [Markdown output](@ref), [Notebook output](@ref) and
+[Script output](@ref). In short, the following is happening:
+
+* Markdown output: markdown chunks are printed as-is, code chunks are put inside
+  a code fence (defaults to `@example`-blocks),
+* Notebook output: markdown chunks are printed in markdown cells, code chunks are
+  put in code cells,
+* Script output: markdown chunks are discarded, code chunks are printed as-is.
+
+
+## [**3.4.** Post-processing](@id Post-processing)
+
+When the document is generated the user, again, has the option to hook-into the generation
+with a custom post-processing function. The reason is that one might want to change
+things that are only visible in the rendered document.
+See [Custom pre- and post-processing](@ref Custom-pre-and-post-processing).
+
+
+## [**3.5.** Writing to file](@id Writing-to-file)
+
+The last step of the generation is writing to file. The result is written to
+`$(outputdir)/$(name)(.md|.ipynb|.jl)` where `outputdir` is the output directory supplied
+by the user (for example `docs/generated`), and `name` is a user supplied filename.
+It is recommended to add the output directory to `.gitignore` since the idea is that
+the generated documents will be generated as part of the build process rather than
+beeing files in the repo.
diff --git a/examples/example.jl b/examples/example.jl
new file mode 100644
index 0000000..0c7f772
--- /dev/null
+++ b/examples/example.jl
@@ -0,0 +1,36 @@
+#' # **7.** Example
+#'
+#' *Output generated with Examples.jl based on
+#' [this](../../../examples/example.jl) source file.*
+#'
+#' This is an example source file for input to Examples.jl.
+#'
+#md #' If you are reading this you are seeing the markdown output
+#md #' generated from the source file, here you can see the corresponding
+#md #' notebook output: [example.ipynb](./example.ipynb)
+#nb #' If you are reading this you are seeing the notebook output
+#nb #' generated from the source file, here you can see the corresponding
+#nb #' markdown output: [example.md](./example.md)
+
+#' ## Rational numbers in Julia
+#' Rational number in julia can be constructed with the `//` operator:
+
+x = 1//3
+y = 2//5
+
+#' Operations with rational number returns a new rational number
+
+x + y
+
+#-
+
+x * y
+
+#' Everytime a rational number is constructed, it will be simplified
+#' using the `gcd` function, for example `2//4` simplifies to `1//2`:
+
+2//4
+
+#' and `2//4 + 2//4` simplifies to `1//1`:
+
+2//4 + 2//4
diff --git a/src/Examples.jl b/src/Examples.jl
index e30de1d..43b08f1 100644
--- a/src/Examples.jl
+++ b/src/Examples.jl
@@ -93,7 +93,14 @@ filename(str) = first(splitext(last(splitdir(str))))
 """
     Examples.script(inputfile, outputdir; kwargs...)
 
-Create a script file.
+Generate a plain script file from `inputfile` and write the result to `outputdir`.
+
+Keyword arguments:
+- `name`: name of the output file, excluding `.jl`. Defaults to the
+  filename of `inputfile`.
+- `preprocess`, `postprocess`: custom pre- and post-processing functions,
+  see the [Custom pre- and post-processing](@ref Custom-pre-and-post-processing)
+  section of the manual. Defaults to `identity`.
 """
 function script(inputfile, outputdir; preprocess = identity, postprocess = identity,
                 name = filename(inputfile), kwargs...)
@@ -103,18 +110,18 @@ function script(inputfile, outputdir; preprocess = identity, postprocess = ident
     @info "generating plain script file from $(inputfile)"
     # read content
     content = read(inputfile, String)
+    # - normalize line endings
+    content = replace(content, "\r\n" => "\n")
 
     # run custom pre-processing from user
     content = preprocess(content)
 
     # run built in pre-processing:
-    ## - normalize line endings
     ## - remove #md lines
     ## - remove #nb lines
     ## - remove leading and trailing #jl
     ## - replace @__NAME__
     for repl in Pair{Any,Any}[
-                    "\r\n" => "\n",
                     r"^#md.*\n?"m => "",
                     r"^#nb.*\n?"m => "",
                     r"^#jl "m => "",
@@ -152,10 +159,25 @@ end
 """
     Examples.markdown(inputfile, outputdir; kwargs...)
 
-Generate a markdown file from the `input` file and write the result to the `output` file.
+Generate a markdown file from `inputfile` and write the result
+to the directory`outputdir`.
+
+Keyword arguments:
+- `name`: name of the output file, excluding `.md`. `name` is also used to name
+  all the `@example` blocks. Defaults to the filename of `inputfile`.
+- `preprocess`, `postprocess`: custom pre- and post-processing functions,
+  see the [Custom pre- and post-processing](@ref Custom-pre-and-post-processing)
+  section of the manual. Defaults to `identity`.
+- `codefence`: A `Pair` of opening and closing code fence. Defaults to
+  ````
+  "```@example \$(name)" => "```"
+  ````
+- `documenter`: boolean that says if the output is intended to use with Documenter.jl.
+  Defaults to `false`. See the the manual section on
+  [Interaction with Documenter](@ref Interaction-with-Documenter).
 """
 function markdown(inputfile, outputdir; preprocess = identity, postprocess = identity,
-                  name = filename(inputfile),
+                  name = filename(inputfile), documenter::Bool = false,
                   codefence::Pair = "```@example $(name)" => "```", kwargs...)
     # normalize paths
     inputfile = realpath(abspath(inputfile))
@@ -163,18 +185,18 @@ function markdown(inputfile, outputdir; preprocess = identity, postprocess = ide
     @info "generating markdown page from $(inputfile)"
     # read content
     content = read(inputfile, String)
+    # - normalize line endings
+    content = replace(content, "\r\n" => "\n")
 
     # run custom pre-processing from user
     content = preprocess(content)
 
     # run built in pre-processing:
-    ## - normalize line endings
     ## - remove #nb lines
     ## - remove leading and trailing #jl lines
     ## - remove leading #md
     ## - replace @__NAME__
     for repl in Pair{Any,Any}[
-                    "\r\n" => "\n",
                     r"^#nb.*\n?"m => "",
                     r"^#jl.*\n?"m => "",
                     r".*#jl$\n?"m => "",
@@ -227,9 +249,21 @@ const JUPYTER_VERSION = v"4.3.0"
     Examples.notebook(inputfile, outputdir; kwargs...)
 
 Generate a notebook from `inputfile` and write the result to `outputdir`.
+
+Keyword arguments:
+- `name`: name of the output file, excluding `.ipynb`. Defaults to the
+  filename of `inputfile`.
+- `preprocess`, `postprocess`: custom pre- and post-processing functions,
+  see the [Custom pre- and post-processing](@ref Custom-pre-and-post-processing)
+  section of the manual. Defaults to `identity`.
+- `execute`: a boolean deciding if the generated notebook should also
+  be executed or not. Defaults to `false`.
+- `documenter`: boolean that says if the source contains Documenter.jl specific things
+  to filter out during notebook generation. Defaults to `false`. See the the manual
+  section on [Interaction with Documenter](@ref Interaction-with-Documenter).
 """
 function notebook(inputfile, outputdir; preprocess = identity, postprocess = identity,
-                  execute::Bool=false,
+                  execute::Bool=false, documenter::Bool = false,
                   name = filename(inputfile), kwargs...)
     # normalize paths
     inputfile = realpath(abspath(inputfile))
@@ -237,19 +271,19 @@ function notebook(inputfile, outputdir; preprocess = identity, postprocess = ide
     @info "generating notebook from $(inputfile)"
     # read content
     content = read(inputfile, String)
+    # normalize line endings
+    content = replace(content, "\r\n" => "\n")
 
     # run custom pre-processing from user
     content = preprocess(content)
 
     # run built in pre-processing:
-    ## - normalize line endings
     ## - remove #md lines
     ## - remove leading and trailing #jl lines
     ## - remove leading #nb
     ## - replace @__NAME__
     ## - replace ```math ... ``` with \begin{equation} ... \end{equation}
     for repl in Pair{Any,Any}[
-                    "\r\n" => "\n",
                     r"^#md.*\n?"m => "",
                     r"^#jl.*\n?"m => "",
                     r".*#jl$\n?"m => "",
diff --git a/test/runtests.jl b/test/runtests.jl
index 69ca7ab..1e1a0ff 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -2,4 +2,4 @@ using Examples
 using Base.Test
 
 # write your own tests here
-@test 1 == 2
+@test 1 == 1