diff --git a/TECHNICAL_DOCS.md b/TECHNICAL_DOCS.md
new file mode 100644
index 0000000..3ecd57c
--- /dev/null
+++ b/TECHNICAL_DOCS.md
@@ -0,0 +1,61 @@
+# How the file system work
+
+# All class of ZipponDB
+
+## Tokenizer
+
+All tokenizer are inspiered (highly ;) ) by the Zig tokenizer. https://github.com/ziglang/zig/blob/master/lib/std/zig/tokenizer.zig
+
+Each tokenizer have it's own token. Each token have multiple tag in the form of an enum. For example the CLI Token can have all those tag:
+
+```zig
+pub const Tag = enum {
+  eof,
+  invalid,
+
+  keyword_run,
+  keyword_help,
+  keyword_describe,
+  keyword_schema,
+  keyword_build,
+  keyword_quit,
+
+  string_literal,
+  identifier,
+};
+```
+
+The tokenizer take a buffer as an array of character (u8) when init, it is the text to tokenize.  
+Then a unique function next, that will iterate over each characther of the buffer and find the next token. The token can be invalid and is then return.
+
+### ZiQL
+
+The ZiQL tokenizer is the main tokenizer. It is the one that tokenize query.
+
+### CLI
+
+The CLI tokenizer is the one that take command and do stuff. It is the simplier as it only have keywords like run, help and describe. As well as string for query (Betwwen "").
+
+### Schema
+
+The schema tokenizer take the .zipponschema file and generate the dtypes.zig file that is then use to build the engine.
+
+## DataEngine
+
+The DataEngine is the class that do stuff with files. Nowhere else should file be manipulate.
+
+## Parser
+
+All Parser take a tokenizer and an allocator and do stuff.
+
+### ZiQL
+
+The ZiQL Parser will take the tokenizer and do stuff depending of the main action (GRAB, DELETE, ADD, UPDATE).
+
+### CLI
+
+The CLI doesnt really have a Parser struct but the main function of the console act like it. Maybe I will do a proper Parser in the future.
+
+### Schema
+
+TODO