diff --git a/Cargo.toml b/Cargo.toml
index 41b900b8..d49d3b56 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -1,4 +1,5 @@
 [workspace]
 members = [
-    "lib"
+    "lib",
+    "python"
 ]
\ No newline at end of file
diff --git a/README.md b/README.md
index 9d48c1f5..33634699 100644
--- a/README.md
+++ b/README.md
@@ -4,7 +4,7 @@ This library is a work in progress of a [RDF](https://www.w3.org/RDF/) stack imp
 
 Its goal is to provide a compliant, safe and fast implementation of W3C specifications in Rust.
 
-The `lib` directory contains the Rust library code.
+The `lib` directory contains the Rust library code and the `python` directory a beginning of Python bindings.
 
 [![Build Status](https://travis-ci.org/Tpt/rudf.svg?branch=master)](https://travis-ci.org/Tpt/rudf)
 [![dependency status](https://deps.rs/repo/github/Tpt/rudf/status.svg)](https://deps.rs/repo/github/Tpt/rudf)
diff --git a/python/Cargo.toml b/python/Cargo.toml
new file mode 100644
index 00000000..524abd07
--- /dev/null
+++ b/python/Cargo.toml
@@ -0,0 +1,19 @@
+[package]
+name = "rudf_python"
+version = "0.1.0"
+authors = ["Tpt <thomas@pellissier-tanon.fr>"]
+license = "MIT/Apache-2.0"
+readme = "../README.md"
+keywords = ["RDF", "N-Triples", "Turtle", "RDF/XML", "SPARQL", "Python"]
+repository = "https://github.com/Tpt/rudf"
+description = """
+Python bindings of the Rudf library
+"""
+
+[lib]
+name = "rudf"
+crate-type = ["cdylib"]
+
+[dependencies]
+rudf = {path = "../lib"}
+cpython = { version = "0.2", features = ["extension-module"]}
diff --git a/python/MANIFEST.in b/python/MANIFEST.in
new file mode 100644
index 00000000..83c99e08
--- /dev/null
+++ b/python/MANIFEST.in
@@ -0,0 +1,2 @@
+include Cargo.toml
+recursive-include src *
\ No newline at end of file
diff --git a/python/pyproject.toml b/python/pyproject.toml
new file mode 100644
index 00000000..2d3b7557
--- /dev/null
+++ b/python/pyproject.toml
@@ -0,0 +1,2 @@
+[build-system]
+requires = ["setuptools", "wheel", "setuptools-rust"]
\ No newline at end of file
diff --git a/python/rudf/__init__.py b/python/rudf/__init__.py
new file mode 100644
index 00000000..056af597
--- /dev/null
+++ b/python/rudf/__init__.py
@@ -0,0 +1 @@
+from .rudf import *
\ No newline at end of file
diff --git a/python/setup.py b/python/setup.py
new file mode 100644
index 00000000..8811d49b
--- /dev/null
+++ b/python/setup.py
@@ -0,0 +1,15 @@
+from setuptools import setup
+
+try:
+    from setuptools_rust import Binding, RustExtension
+except ImportError:
+    print('You should install the setuptool-rust package to be able to build rudf')
+
+
+setup(
+    name="rudf",
+    version="0.1",
+    rust_extensions=[RustExtension("rudf.rudf", binding=Binding.RustCPython)],
+    packages=["rudf"],
+    zip_safe=False,
+)
\ No newline at end of file
diff --git a/python/src/lib.rs b/python/src/lib.rs
new file mode 100644
index 00000000..9b515459
--- /dev/null
+++ b/python/src/lib.rs
@@ -0,0 +1,128 @@
+#[macro_use]
+extern crate cpython;
+extern crate rudf;
+
+use cpython::exc::ValueError;
+use cpython::CompareOp;
+use cpython::PyErr;
+use cpython::PyResult;
+use cpython::Python;
+use cpython::PythonObject;
+use cpython::ToPyObject;
+use rudf::model;
+use std::collections::hash_map::DefaultHasher;
+use std::error::Error;
+use std::hash::Hash;
+use std::hash::Hasher;
+use std::str::FromStr;
+
+py_module_initializer!(rudf, initrudf, PyInit_rudf, |py, m| {
+    try!(m.add(py, "__doc__", "Rudf Python bindings"));
+    try!(m.add_class::<NamedNode>(py));
+    try!(m.add_class::<BlankNode>(py));
+    try!(m.add_class::<Literal>(py));
+    Ok(())
+});
+
+fn new_value_error(py: Python, error: impl Error) -> PyErr {
+    PyErr::new::<ValueError, _>(py, error.description())
+}
+
+fn eq_compare<T: Eq + Ord>(a: T, b: T, op: CompareOp) -> bool {
+    match op {
+        CompareOp::Lt => a < b,
+        CompareOp::Le => a <= b,
+        CompareOp::Eq => a == b,
+        CompareOp::Ne => a != b,
+        CompareOp::Gt => a > b,
+        CompareOp::Ge => a >= b,
+    }
+}
+
+fn hash(t: &impl Hash) -> u64 {
+    let mut s = DefaultHasher::new();
+    t.hash(&mut s);
+    s.finish()
+}
+
+py_class!(class NamedNode |py| {
+    data inner: model::NamedNode;
+
+    def __new__(_cls, value: &str) -> PyResult<NamedNode> {
+        NamedNode::create_instance(py, model::NamedNode::from_str(value).map_err(|error| new_value_error(py, error))?)
+    }
+
+    def value(&self) -> PyResult<String> {
+        Ok(self.inner(py).as_str().to_string())
+    }
+
+    def __str__(&self) -> PyResult<String> {
+        Ok(self.inner(py).to_string())
+    }
+
+    def __richcmp__(&self, other: &NamedNode, op: CompareOp) -> PyResult<bool> {
+        Ok(eq_compare(self.inner(py), other.inner(py), op))
+    }
+
+    def __hash__(&self) -> PyResult<u64> {
+        Ok(hash(self.inner(py)))
+    }
+});
+
+py_class!(class BlankNode |py| {
+    data inner: model::BlankNode;
+
+    def __new__(_cls) -> PyResult<BlankNode> {
+        BlankNode::create_instance(py, model::BlankNode::default())
+    }
+
+    def __str__(&self) -> PyResult<String> {
+        Ok(self.inner(py).to_string())
+    }
+
+    def __richcmp__(&self, other: &BlankNode, op: CompareOp) -> PyResult<bool> {
+        Ok(eq_compare(self.inner(py), other.inner(py), op))
+    }
+
+    def __hash__(&self) -> PyResult<u64> {
+        Ok(hash(self.inner(py)))
+    }
+});
+
+py_class!(class Literal |py| {
+    data inner: model::Literal;
+
+    def __new__(_cls, value: String, language: Option<String> = None, datatype: Option<NamedNode> = None) -> PyResult<Literal> {
+        Literal::create_instance(py, match language {
+            Some(language) => model::Literal::new_language_tagged_literal(value, language),
+            None => match datatype {
+                Some(datatype) => model::Literal::new_typed_literal(value, datatype.inner(py).clone()),
+                None => model::Literal::new_simple_literal(value)
+            }
+        })
+    }
+
+    def value(&self) -> PyResult<String> {
+        Ok(self.inner(py).value().to_string())
+    }
+
+    def language(&self) -> PyResult<Option<String>> {
+        Ok(self.inner(py).language().map(|l| l.to_string()))
+    }
+
+    def datatype(&self) -> PyResult<NamedNode> {
+        NamedNode::create_instance(py, self.inner(py).datatype().clone())
+    }
+
+    def __str__(&self) -> PyResult<String> {
+        Ok(self.inner(py).to_string())
+    }
+
+    def __richcmp__(&self, other: &Literal, op: CompareOp) -> PyResult<bool> {
+        Ok(eq_compare(self.inner(py), other.inner(py), op))
+    }
+
+    def __hash__(&self) -> PyResult<u64> {
+        Ok(hash(self.inner(py)))
+    }
+});
diff --git a/python/tests/test_model.py b/python/tests/test_model.py
new file mode 100644
index 00000000..3172dea0
--- /dev/null
+++ b/python/tests/test_model.py
@@ -0,0 +1,33 @@
+import unittest
+from rudf import *
+
+XSD_STRING = NamedNode('http://www.w3.org/2001/XMLSchema#string')
+XSD_INTEGER = NamedNode('http://www.w3.org/2001/XMLSchema#integer')
+RDF_LANG_STRING = NamedNode('http://www.w3.org/1999/02/22-rdf-syntax-ns#langString')
+
+
+class TestNamedNode(unittest.TestCase):
+    def test_constructor(self):
+        self.assertEqual(NamedNode('http://foo').value(), 'http://foo/')
+
+
+class TestBlankNode(unittest.TestCase):
+    def test_constructor(self):
+        self.assertNotEqual(BlankNode(), BlankNode())
+
+
+class TestLiteral(unittest.TestCase):
+    def test_constructor(self):
+        self.assertEqual(Literal('foo').value(), 'foo')
+        self.assertEqual(Literal('foo').datatype(), XSD_STRING)
+
+        self.assertEqual(Literal('foo', 'en').value(), 'foo')
+        self.assertEqual(Literal('foo', 'en').language(), 'en')
+        self.assertEqual(Literal('foo', 'en').datatype(), RDF_LANG_STRING)
+
+        self.assertEqual(Literal('foo', datatype=XSD_INTEGER).value(), 'foo')
+        self.assertEqual(Literal('foo', datatype=XSD_INTEGER).datatype(), XSD_INTEGER)
+
+
+if __name__ == '__main__':
+    unittest.main()