diff --git a/docs/Pipeline.md b/docs/Pipeline.md
index 454a092fc..9c9781eb7 100644
--- a/docs/Pipeline.md
+++ b/docs/Pipeline.md
@@ -1 +1,23 @@
 # Pipeline
+
+```mermaid
+classDiagram
+    class AbstractPipeline
+    class Pipeline
+    class PipelineStep
+
+    Pipeline--> "*" AbstractPipeline
+    Pipeline..|>AbstractPipeline
+    PipelineStep..|>AbstractPipeline
+```
+
+For the pipeline definition, we use a composite to allow us to have a multi-level pipeline.
+As such, a pipeline consists of an arbitrary number of either *PipelineStep*s or further *Pipeline*s.
+
+In our approach, we use three levels for our pipeline:
+On the first level, the overall pipeline defines multiple *stages*, e.g., text preprocessing or element connection.
+Each stage is another pipeline that then defines *agents* that have the purpose of initiating the processing and of collecting the information of the various heuristics.
+Agents then use *Informants* as concrete PipelineSteps to execute the processing and heuristics.
+
+A pipeline step (i.e., an Informant) stores results within a repository that can be universally accessed by all pipeline steps, similarly to a blackboard in the blackboard pattern.
+This way, each pipeline step and, thus, each heuristic can access the results of previous steps and provide its results for others.