diff --git a/tesseract_ros_examples/examples_readme/puzzle_aux.md b/tesseract_ros_examples/examples_readme/puzzle_aux.md
new file mode 100644
index 000000000..2d45605ce
--- /dev/null
+++ b/tesseract_ros_examples/examples_readme/puzzle_aux.md
@@ -0,0 +1,129 @@
+# Puzzle Piece auxiliary axes example
+
+The puzzle piece examples show a small collaborative robot manipulating a puzzle piece to debur the edges with a grinder which has two axes of rotation.
+
+![Puzzle piece](https://github.com/ros-industrial-consortium/trajopt_ros/blob/master/gh_pages/_static/example_gifs/puzzle_piece_with_positioner.gif)
+
+```cpp
+tesseract_common::VectorIsometry3d tool_poses = makePuzzleToolPoses();
+```
+
+`makePuzzleToolPoses`: reads the target poses of the puzzle piece from the
+`/config/puzzle_bent.csv` and return a list of the target poses.
+
+```cpp
+// Create cartesian waypoint
+Waypoint wp = CartesianWaypoint(tool_poses[0]);
+PlanInstruction plan_instruction(wp, PlanInstructionType::START, "CARTESIAN");
+plan_instruction.setDescription("from_start_plan");
+program.setStartInstruction(plan_instruction);
+```
+
+The first pose is set as the start poses.
+
+```cpp
+for (std::size_t i = 1; i < tool_poses.size(); ++i)
+{
+    Waypoint wp = CartesianWaypoint(tool_poses[i]);
+    PlanInstruction plan_instruction(wp, PlanInstructionType::LINEAR, "CARTESIAN");
+    plan_instruction.setDescription("waypoint_" + std::to_string(i));
+    program.push_back(plan_instruction);
+}
+```
+
+A linear motion is set between the rest of the points.
+
+```cpp
+const std::string new_planner_name = "TRAJOPT_NO_POST_CHECK";
+tesseract_planning::TrajOptTaskflowParams params;
+params.enable_post_contact_discrete_check = false;
+params.enable_post_contact_continuous_check = false;
+planning_server.registerProcessPlanner(new_planner_name,
+                                        std::make_unique<tesseract_planning::TrajOptTaskflow>(params));
+```
+
+A trajopt taskflow is used without collision checking by setting
+`enable_post_contact_discrete_check` & `enable_post_contact_continuous_check` to `false`.
+
+Next we will create TrajOpt Profiles.
+
+```cpp
+// Create TrajOpt Profile
+auto trajopt_plan_profile = std::make_shared<tesseract_planning::TrajOptDefaultPlanProfile>();
+trajopt_plan_profile->cartesian_coeff = Eigen::VectorXd::Constant(6, 1, 5);
+trajopt_plan_profile->cartesian_coeff(3) = 2;
+trajopt_plan_profile->cartesian_coeff(4) = 2;
+trajopt_plan_profile->cartesian_coeff(5) = 0;
+```
+
+In the `TrajOptDefaultPlanProfile`, the `cartesian_coeff` is a vector of size six.
+The first three correspond to translation and the last three to rotation.
+The `trajopt_plan_profile->cartesian_coeff(5) = 0;` indicates it is free to rotate around the z-axis.
+If the value is greater than zero it is considered constrained and the coefficient represents a
+weight/scale applied to the error and gradient.
+
+```cpp
+auto trajopt_composite_profile = std::make_shared<tesseract_planning::TrajOptDefaultCompositeProfile>();
+trajopt_composite_profile->collision_constraint_config.enabled = false;
+trajopt_composite_profile->collision_cost_config.enabled = true;
+trajopt_composite_profile->collision_cost_config.safety_margin = 0.025;
+trajopt_composite_profile->collision_cost_config.type = trajopt::CollisionEvaluatorType::SINGLE_TIMESTEP;
+trajopt_composite_profile->collision_cost_config.coeff = 1;
+```
+
+In the `TrajOptDefaultCompositeProfile`, we can set the following parameters:
+
+`collision_constraint_config`: to decide to apply constraints on the collision
+
+`collision_cost_config`: to give costs for the collision.
+
+`collision_cost_config.safety_margin`: define the safety margin for collision
+
+`collision_cost_config.type`: It can take the following values:
+
+* SINGLE_TIMESTEP: TODO, description.
+* DISCRETE_CONTINUOUS: TODO, description.
+* CAST_CONTINUOUS: TODO, description.
+
+```cpp
+auto trajopt_solver_profile = std::make_shared<tesseract_planning::TrajOptDefaultSolverProfile>();
+trajopt_solver_profile->convex_solver = sco::ModelType::OSQP;
+trajopt_solver_profile->opt_info.max_iter = 200;
+trajopt_solver_profile->opt_info.min_approx_improve = 1e-3;
+trajopt_solver_profile->opt_info.min_trust_box_size = 1e-3;
+```
+
+In `TrajOptDefaultSolverProfile` we specify the type of the convex solver to use it can be:
+GUROBI, OSQP, QPOASES, BPMPD, AUTO_SOLVER
+
+`max_iter`: the maximum number of iterations to find a solution.
+
+`min_approx_improve`: the minimum improvement that for the solution after each iteration.
+
+`min_trust_box_size`: TODO, describe
+
+```cpp
+// Add profile to Dictionary
+planning_server.getProfiles()->addProfile<tesseract_planning::TrajOptPlanProfile>("CARTESIAN",          trajopt_plan_profile);
+planning_server.getProfiles()->addProfile<tesseract_planning::TrajOptCompositeProfile>("DEFAULT",
+                                                                                        trajopt_composite_profile);
+planning_server.getProfiles()->addProfile<tesseract_planning::TrajOptSolverProfile>("DEFAULT",
+                                                                                    trajopt_solver_profile);
+```
+
+We add the profiles created above to our planning server.
+
+```cpp
+// Create Process Planning Request
+ProcessPlanningRequest request;
+request.name = new_planner_name;
+request.instructions = Instruction(program);
+```
+
+We create a request object and add the program to its instruction.
+
+```cpp
+ProcessPlanningFuture response = planning_server.run(request);
+```
+
+we use the run method providing the request as a parameter to get the response.