diff --git a/.travis.yml b/.travis.yml
index 8ca0083b..d8c711ef 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -15,6 +15,18 @@ matrix:
             - bash ./scripts/license-headers-updated
             - bash ./scripts/branch-contains-no-tmp-commits
             - bash ./scripts/version-updated
+        - language: rust
+          rust: stable
+          name: userdoc
+          cache:
+            directories:
+              - /home/travis/.cargo
+          before_cache:
+            - rm -rf /home/travis/.cargo/registry
+          install:
+            - cargo install mdbook --force
+          script:
+            - cd doc/user && mdbook build || exit 1
         - language: rust
           rust: 1.35.0
           cache:
diff --git a/doc/user/.gitignore b/doc/user/.gitignore
new file mode 100644
index 00000000..7585238e
--- /dev/null
+++ b/doc/user/.gitignore
@@ -0,0 +1 @@
+book
diff --git a/doc/user/book.toml b/doc/user/book.toml
new file mode 100644
index 00000000..c866c6e4
--- /dev/null
+++ b/doc/user/book.toml
@@ -0,0 +1,6 @@
+[book]
+authors      = ["Matthias Beyer"]
+language     = "en"
+multilingual = false
+src          = "src"
+title        = "imag-documentation"
diff --git a/doc/user/src/SUMMARY.md b/doc/user/src/SUMMARY.md
new file mode 100644
index 00000000..1dd182a8
--- /dev/null
+++ b/doc/user/src/SUMMARY.md
@@ -0,0 +1,8 @@
+# Summary
+
+- [Introduction](./intro.md)
+- [Approach](./approach.md)
+- [Conventions](./conventions.md)
+- [Good to know](./good-to-know.md)
+- [Examples](./examples.md)
+
diff --git a/doc/user/src/approach.md b/doc/user/src/approach.md
new file mode 100644
index 00000000..ae69b997
--- /dev/null
+++ b/doc/user/src/approach.md
@@ -0,0 +1,74 @@
+# The Approach
+
+The approach "imag" takes on solving this problem is to store content in a
+plain text storage (the "store") on the filesystem and persisting content in a
+unified way.
+
+## The Store
+
+The imag "store" is nothing more than a directory on the filesystem, ususally
+`~/.imag/store`.
+Imag stores each "entry" under a unique "StoreId" which is nothing more than a
+part of the path of the actual file on disk:
+
+
+```
+/home/user/.imag/store/some/entry
+|               |     ||        |
+ \             /     /  \      /
+  -------------      |   ------
+|   "RTP"            |    "Id"
+ \                  /
+  ------------------
+    "Store path"
+```
+
+* The "RTP" is the Runtimepath of imag, where imag looks for the config file
+  (`~/.imag/imagrc.toml`) and the `store` directory.
+
+* The "Store path" is where imag looks for the store root. The "Store path" can
+  be set explicitely in the imag commandline, but end-users normally don't need
+  to do this.
+
+* The "StoreId" (abbreviated with "Id") is what the user uses when referring to
+  an entry of the store.
+
+
+## Entry
+
+One imag entry contains two parts: The Header and the Content part. The header,
+which is stored in
+[TOML](https://en.wikipedia.org/wiki/TOML),
+contains structured data. This data is most of the time generated by imag
+itself, depending on the module that stores the data.
+An imag entry contains always a version information header part:
+
+```
+[imag]
+version = "0.10.0"
+```
+
+Each imag module is free to store data under its own header section, where the
+name of the section is defined by the module storing it. Thus, "imag-diary"
+stores its data in `[diary]`.
+A module may store arbitrary textual data.
+Of course, imag provides utility commandline tools for querying this data.
+
+An example imag entry could look like this:
+
+```
+---
+[habit.instance]
+comment = 'Eat one fruit per day'
+date = '2019-06-28'
+is_habit_instance = true
+name = 'eat_fruit'
+
+[imag]
+version = '0.10.0'
+
+[links]
+internal = ['habit/template/eat_fruit']
+---
+```
+
diff --git a/doc/user/src/conventions.md b/doc/user/src/conventions.md
new file mode 100644
index 00000000..2bf61acf
--- /dev/null
+++ b/doc/user/src/conventions.md
@@ -0,0 +1,46 @@
+# Conventions
+
+imag has a few conventions which we try to enforce in all module
+implementations. This chapter explains the conventions an end-user of imag can
+rely on.
+
+
+## Commandline interface
+
+The commandline interface of every subcommand has a "--help" flag which can be
+used to print helptext for the subcommand.
+
+Every imag module implementation can be used for scripting, where it behaves in
+the following ways:
+
+* If the standard input is a pipe, imag module implementations assume that store
+  ids are written to that pipe line by line.
+* If the standard output is a pipe, every imag module prints the StoreIDs it
+  touched while running to the pipe.
+  All other output is written to stderr.
+  Piping can be used to combine imag commands.
+* If the standard output is not a pipe, the imag module does not print the
+  StoreIDs it touched.
+
+This behaviour can be overridden with the `--ignore-ids` flag.
+
+## Versioning
+
+imag modules are compatible to eachother as long as the version number is in the
+`0.x.y` range.
+Modules from different imag versions are not supported to be compatible to
+eachother.
+
+
+## Commandline capabilities
+
+The imag commands can be used to access _all_ data that is stored in the imag
+store.
+Alternatively, standard unix commandline tools (like `grep`, `cut`, `sed`, ...)
+can be used to access all data and all data-points.
+
+In short: All data imag stores is stored in plain text, containing a structured
+part (in the markup language "toml") and a plain-text part (which should be
+UTF-8 encoded).
+
+
diff --git a/doc/user/src/examples.md b/doc/user/src/examples.md
new file mode 100644
index 00000000..fd6e57c6
--- /dev/null
+++ b/doc/user/src/examples.md
@@ -0,0 +1,22 @@
+# Examples
+
+imag commands can be chained by piping the output.
+This can be used to create, tag and categorize an entry in one step.
+
+```
+imag log --to personal "Some personal note" | \
+    imag tag add foobar | \
+    imag category set somecategory
+```
+
+imag can be configured to use aliases for module commands, so combining a basi
+alias `alias i=imag` with imag module aliaes, this can be condensed to:
+
+```
+i l --to personal "Some personal note" | \
+    i t add foobar | \
+    i c set somecategory
+```
+
+for example.
+
diff --git a/doc/user/src/good-to-know.md b/doc/user/src/good-to-know.md
new file mode 100644
index 00000000..70bdaae5
--- /dev/null
+++ b/doc/user/src/good-to-know.md
@@ -0,0 +1,34 @@
+# Good to know
+
+There are some things in the imag space that a user might want to know before
+using imag.
+
+
+## Tags
+
+Tags are text values that can be added to _any_ entry in the imag store.
+The tag has to satisfy some invariants to be valid:
+
+1. it has to be all-lowercase
+1. it has no spaces
+1. all characters are alphanumeric
+
+Any entry can have any number of tags. Spelling is not checked.
+
+Finding all entries for a tag "foo" is `O(n)` where `n` is the number of entries
+in the store.
+
+
+## Categories
+
+Categories are more restrictive, but also more powerful.
+First of all, a category has to exist before an entry can be _linked_ to a
+category. A category therefor is represented by an entry and if an entry has a
+category, it is linked to said entry.
+
+So, if you create only category with the name "foo", you cannot set the category
+"bar" for some entry.
+
+Because categories are linked, finding all entries for a category is trivial and
+a `O(1)` operation.
+
diff --git a/doc/user/src/intro.md b/doc/user/src/intro.md
new file mode 100644
index 00000000..f474673a
--- /dev/null
+++ b/doc/user/src/intro.md
@@ -0,0 +1,15 @@
+# Introduction
+
+This document is the end-user documentation for imag, the plain-text personal
+information management suite for the commandline.
+
+The imag project wants to provide a set of tools (called "modules") for managing
+personal information data on the commandline, where the data is stored in plain
+text.
+The modules should be scriptable, composable, traverseable and queryable.
+The target audience for imag are power-users and commandline natives.
+
+If you have any objections, suggestions for improvements, bugs, etc, please file
+them.
+
+