diff --git a/README.md b/README.md
index d3ea695..dbda66b 100644
--- a/README.md
+++ b/README.md
@@ -9,5 +9,32 @@ Tags are exported in a yaml file. Each key corresponds to a tag in the messy
 set, and contains a list of tags of the new set to which this tag should be
 merged.
 
-The script can take the yaml file back as input, which allows pausing and
-restarting the tag merging process.
+## Usage
+
+To start merging a set, with a file containing one tag per line:
+
+```
+tagmerger --export-to <output yaml file> --plain-input <your input file>
+```
+
+Start merging with the clunky interface :p
+When you are bored, hit ^C. The merge progress have been saved to the specified
+yaml file.
+
+To resume the merge:
+
+```
+tagmerger --export-to <output yaml file> <input yaml file>
+```
+## Caveats
+
+Very much work in progress. The interface is very verbose and quite
+impractical, but it can still help a bit merging large sets where it is not
+feasable to keep all tags in human memory.
+
+One of the main issues is that there is no way to revert a merge from the
+interface. If you accidently made a bad merge, stop tagmerger, open the yaml
+output file, fix the merge manually, and resume the merge with the yaml file as
+input.
+
+I think a more graphical interface would make all this much easier...