diff --git a/prototyping/structure.md b/prototyping/structure.md
new file mode 100644
index 0000000..15c820f
--- /dev/null
+++ b/prototyping/structure.md
@@ -0,0 +1,80 @@
+
+
+Abratest has 3 required inputs, but most importantly the test configuration is done through the .env files given with the --env_paths argument. So let's say we want to run abratest with these 3 .env files:
+
+- config1.env [of TYPE authentik]
+
+- config2.env [of TYPE wordpress]
+
+- config3.env [of TYPE wordpress]
+
+Now we run 
+
+```bash
+abratest --env_paths path/config1.env;path/config2.env;path/config3.env [...other args]
+```
+
+
+```
+abratest -> create Coordinator() instance
+└── Coordinator() -> create Runner() subclass instances
+    ├── RunnerAuthentik() [based on config1.env, loaded 
+    │   │                  from abra/recipes/authentik]
+    │   │     # RunnerAuthentik with 3 test files:
+    │   ├── RUN pytest path/setup_authentik.py
+    │   ├── RUN pytest path/test_authentik_1.py
+    │   └── RUN pytest path/test_authentik_2.py
+    ├── RunnerWordpress() [based on config2.env, loaded
+    │   │                  from abra/recipes/wordpress]
+    │   │     # RunnerWordpress with 1 test file
+    │   ├── RUN pytest path/setup_authentik.py
+    │   ├── RUN pytest path/test_authentik_1.py
+    │   └── RUN pytest path/test_authentik_2.py
+    └── RunnerWordpress() [based on config3.env, loaded
+        │                  from abra/recipes/wordpress]
+        │     # RunnerWordpress with 1 test file
+        ├── RUN pytest path/setup_authentik.py
+        ├── RUN pytest path/test_authentik_1.py
+        └── RUN pytest path/test_authentik_2.py
+
+
+```
+
+Coordinator will take care of the correct order of the tests. In general, tests are placed in one of 3 categories: `setups`, `tests` and `cleanups`. To associate a test with one of these categories, place the Test in the corresponding list of the Runner class, i.e. Runner.setups = [test] or Runner.tests = [test]. The execution order will be.
+
+> [setups] ➔ [tests] ➔ [cleanups]
+
+
+Furthermore, some `Runner` classes can depend on others. For example, `RunnerWordpress` depends on `RunnerAuthentik`. Therefore, `Coordinator` will make sure that `RunnerAuthentik` runs before `RunnerWordpress`. We will end up with with this order:
+
+| #   | Runner         | Type     |
+| --- | -------------- | -------- |
+| 1.  | Authentik      | setups   |
+| 2.  | Wordpress-1    | setups   |
+| 3.  | Wordpress-2    | setups   |
+| 4.  | Authentik      | tests    |
+| 5.  | Wordpress-1    | tests    |
+| 6.  | Wordpress-2    | tests    |
+| 7.  | Authentik      | cleanups |
+| 8.  | Wordpress-1    | cleanups |
+| 9.  | Wordpress-2    | cleanups |
+
+
+
+To comprehend this process, let's examine a simplified rendition of the `RunnerWordpress` class. Within it, there exist two setup scripts and two test scripts, one of which operates conditionally.
+
+
+```python
+class RunnerWordpress(Runner):
+    env_type = "wordpress"
+    dependencies = ["authentik"]
+    setups = [
+        Test(test_file="setup_wordpress_1.py"),
+        Test(test_file="setup_wordpress_2.py"),
+    ]
+    tests = [
+        Test(test_file="test_wordpress.py"),
+        Test(condition=condition_function, test_file="test_wordpress_conditional.py"),
+    ]
+    cleanups = []
+```