We have been building Sancho in relative quiet since the spring. A few months of prototypes, a few early pilot conversations, a lot of time on factory floors watching the same integration problem appear in different cells at different plants. Today we are ready to talk about what we built and why.
The short version: Sancho is an orchestration layer that sits between your robot controllers and your shop floor systems. It speaks FANUC, KUKA, ABB, Universal Robots, and Siemens in their native protocols, and it presents a single, consistent interface to everything above it. One cell map configuration. One error propagation model. One place to change when the process changes.
The Problem We Kept Seeing
Rafael and I both came from backgrounds in industrial automation before we started this. Between us we have spent a combined decade or so commissioning production cells: electronics assembly, automotive body shops, welding lines, material handling. The cells varied enormously, but one thing was constant: every time a cell had more than one robot vendor in it, the integration was built from scratch.
Not slightly adapted from a previous project. From scratch. New ladder logic for the PLC-mediated handoffs. New custom scripts for the cross-vendor cycle timing. New documentation that described how this specific cell's specific wiring mapped to the output bits of these specific controllers. And when something changed, when the plant added a cobot to an existing FANUC line, when a UR10 replaced a KUKA that had reached end-of-life, the integration work started over.
We started mapping this problem more carefully when we were both at the same automation systems firm. The total labor cost for initial integration of a two-vendor cell was predictable. What was less visible was the maintenance tail: the ongoing cost of keeping that custom integration working through firmware updates, process changes, and personnel turnover. We started tracking it and found that the maintenance cost over three years often equaled or exceeded the initial build cost. That number surprised even people who had been doing this work for years.
What We Built
Sancho has three layers. The connector layer provides protocol-native interfaces to each supported vendor. The FANUC connectors use the FANUC Robot Interface protocol and FOCAS2 for CNC machines. The KUKA connectors use RSI/XML. The ABB connectors use RAPID sockets and EGM. The UR connector uses RTDE. The Siemens connector uses S7comm. Each connector handles the vendor-specific protocol quirks and normalizes the data into a common internal representation.
The orchestration engine coordinates across connectors. It manages the cell lifecycle: startup sequencing, handoff coordination between robots, error propagation when any device faults, shutdown sequencing. The cell map is a configuration file, not code, that declares what devices are in the cell, what their network addresses are, and what the coordination rules are between them. Adding a new device to an existing cell means adding an entry to the cell map; it does not mean rewriting the coordination logic.
The interface layer presents the orchestration state upward to whatever shop floor system needs it. Sancho can publish to MQTT, OPC-UA, or a REST API. The MES does not need to know which robot vendor is running in each cell; it interacts with a normalized cell state representation that is the same regardless of what hardware is underneath.
Who Should Talk to Us
Right now we are working with a small number of early-access partners who have mixed-vendor cells and are willing to commission Sancho alongside us. We are not trying to be everything to everyone at this stage. The specific profile we are looking for: plants with two or more robot vendors in one production cell, where the integration is currently custom-built and the maintenance overhead is noticeable. Automotive suppliers, electronics contract manufacturers, and industrial equipment producers tend to fit this description.
We are also interested in talking to automation system integrators who are building cells for their customers and are tired of writing the same cross-vendor integration boilerplate on every project. Sancho can be part of their standard cell architecture rather than a custom build each time.
We are not the right fit for plants with single-vendor standardized cells where the integration overhead is low, or for applications that need sub-millisecond coordination between robots (we are in the single-digit milliseconds range for cross-vendor coordination events, which covers most production cell applications but not high-speed precision assembly).
Where We Are With Protocol Coverage
As of this launch, the connectors we are comfortable putting in a production cell are: FANUC robot interface (R-30iB, R-30iB Plus, R-30iB Mate controllers), KUKA RSI for KRC4-based controllers, ABB RAPID Socket and EGM for IRC5 and OmniCore, Universal Robots RTDE for e-Series, Siemens S7-1200 and S7-1500 via S7comm, and Modbus TCP as a generic I/O bridge. These cover the majority of cells we have encountered in early conversations.
We have FANUC CNC (FOCAS2) in beta and EtherNet/IP in beta. Mitsubishi MELFA is on the roadmap for Q1 2026. If your cell has hardware that is not on this list, reach out. We have added connectors quickly when there is a clear pilot application, and the connector architecture is designed to make new vendors tractable in days rather than months.
Read next: The Anatomy of a Mixed-Vendor Workcell: What Each Layer Actually Does