diff --git a/gym_dora/gym_dora/env.py b/gym_dora/gym_dora/env.py
index 928ef27c..d0b406e3 100644
--- a/gym_dora/gym_dora/env.py
+++ b/gym_dora/gym_dora/env.py
@@ -1,53 +1,55 @@
 import gymnasium as gym
 import numpy as np
+from dora import Node
+import os
+import pyarrow as pa
 
+IMAGE_WIDTH = int(os.getenv("IMAGE_WIDTH", "640"))
+IMAGE_HEIGHT = int(os.getenv("IMAGE_HEIGHT", "480"))
+FPS = int(os.getenv("FPS", "30"))
 
 class DoraEnv(gym.Env):
-    metadata = {"render_modes": ["rgb_array"], "render_fps": 30}
+    metadata = {"render_modes": ["rgb_array"], "render_fps": FPS}
 
     def __init__(self, model="aloha"):
-        ...
-        # TODO: add code to connect with dora client here
+        self.node = Node()
+        self.observation = {"pixels": {}, "terminated": False}
+
+    def _update(self) -> dict: 
+        while True:
+            event = self.node.next(timeout=0.001)
+
+            if event is None:
+                self.observation["terminated"] = True
+                break
+            if event["type"] == "INPUT":
+                if "cam" in event["id"]:
+                    self.observation["pixels"][event["id"]] = event["value"].to_numpy().reshape(IMAGE_HEIGHT, IMAGE_WIDTH, 3)
+                else:
+                    self.observation[event["id"]] = event["value"].to_numpy()
+            elif event["type"] == "ERROR":
+                break
+        
 
     def reset(self, seed: int | None = None):
-        ...
-        # TODO: same as `step` but doesn't take `actions`
-        observation = {
-            "pixels": {
-                "top": ...,
-                "bottom": ...,
-                "left": ...,
-                "right": ...,
-            },
-            "agent_pos": ...,
-            # "agent_vel": ...,  # will be added later
-        }
+
+        self._update()
+        
         reward = 0
-        terminated = truncated = False
+        terminated = truncated = self.observation["terminated"]
         info = {}
-        return observation, reward, terminated, truncated, info
+        return self.observation, reward, terminated, truncated, info
 
     def render(self): ...
 
     def step(self, action: np.ndarray):
-        ...
-        # TODO: this is the important bit: the data to be return by Dora to the policy.
-        observation = {
-            "pixels": {
-                "top": ...,
-                "bottom": ...,
-                "left": ...,
-                "right": ...,
-            },
-            "agent_pos": ...,
-            # "agent_vel": ...,  # will be added later
-        }
+        self._update()
+        self.node.send_output("action", pa.array(action))
         reward = 0
-        terminated = truncated = False
+        terminated = truncated = self.observation["terminated"]
         info = {}
-        return observation, reward, terminated, truncated, info
+        return self.observation, reward, terminated, truncated, info
 
     def close(self):
-        pass
-        # TODO: If code needs to be run when closing the env (e.g. shutting down Dora client),
-        # this is the place to do it. Otherwise this can stay as is.
+        del self.node
+
diff --git a/gym_dora/pyproject.toml b/gym_dora/pyproject.toml
index 7cc210f9..d4dc9975 100644
--- a/gym_dora/pyproject.toml
+++ b/gym_dora/pyproject.toml
@@ -4,12 +4,13 @@ version = "0.1.0"
 description = ""
 authors = ["Simon Alibert <alibert.sim@gmail.com>"]
 readme = "README.md"
-packages = [{include = "gym_dora"}]
+packages = [{ include = "gym_dora" }]
 
 [tool.poetry.dependencies]
 python = "^3.10"
 gymnasium = ">=0.29.1"
-
+dora-rs = ">=0.3.4"
+pyarrow = ">=12.0.0"
 
 [build-system]
 requires = ["poetry-core"]