Industrial Connectivity adds machine data, Native MCP Connectors link enterprise systems and AI tools, Composable AI Agents act on that record, and Factory Playback adds what cameras see on the floor.
BOSTON, Mass., October 6, 2026 — Tulip, the leader in Frontline Operations, announced a set of operational intelligence capabilities at Operations Calling 2026 that help manufacturers run production and orchestrate the people, machines, and AI agents on the floor from one operational record. Tulip previewed an Industrial Connectivity offering, which brings machine data onto that record, and Native MCP Connectors, which exchange data in both directions between Tulip and other enterprise systems and AI tools. Composable AI Agents, introduced at Operations Calling 2025, work across that record. Tulip also showed new Factory Playback capabilities, including human activity detection to improve process analysis.
Manufacturing runs in physical space, which makes it slower to change than other software. A single plant can run machines from several decades and vendors, each reporting data its own way, and much of what happens between digital records, such as a material move, a wait at a workstation, or a skipped step, is never captured. AI agents are beginning to take on work alongside frontline teams, but because an Agent's actions on a production floor impact machines and people, manufacturers need a record of what it reasoned on and what it changed before they put it into production.
These capabilities enable operations to more effectively run and orchestrate manufacturing at scale, and to make changes on the floor faster. They put clean machine and enterprise data on one record, put AI agents to work on it, and show what happened on the line, so a team does root cause analysis, test solutions, and confirm results without piecing evidence together from separate systems. Every decision, by a person or an agent, traces back to the machine data, video, and order behind it.
"Shop floors are a mix of different machines, generations, protocols, naming conventions, and units. You can shove it all into one namespace, but if it isn't cleaned and unified first, you're not going to be able to do much with that data. Data harmonization and normalization have become even more important as AI has increased its presence in operations," said Mason Glidden, Chief Product and Engineering Officer at Tulip.
Industrial Connectivity and Native MCP Connectors: machines and enterprise systems on one record
Industrial Connectivity is Tulip's operational data layer. It brings machine data from a plant's existing drivers, historians, and UNS into Tulip and joins it with real-time human context at scale. At the plant, it normalizes that data as it arrives, converting units and deriving machine states from signals, and maps each machine to a standard type in a common data model, so equipment from different vendors and generations reports the same way. Unlike platforms that stop at asset-tag modeling, Tulip ties machine data directly to the order, step, station, and operator, capturing not just what the machine did, but what the human decided and why.
Native MCP Connectors extend that record to the rest of the enterprise. Using the Model Context Protocol (MCP), an open standard for connecting AI tools to business systems, Tulip exchanges data in both directions with enterprise systems and AI tools that support MCP. Tulip Agents can pull data from those systems and act on them, and AI assistants and coding tools can read from and act on Tulip through the Tulip MCP server. Every request runs under the permissions a company sets in Tulip, so an AI tool given read-only access can view records but cannot change them. Native MCP Connectors roll out over the coming months, first in agents and then in automated workflows.
Composable AI Agents: more of the floor, same rules
Composable AI Agents have been available to Tulip customers since Operations Calling 2025, and more than 40% of customers now run at least one agent in production. One specialty vehicle manufacturer uses an agent to investigate defects on its shop floor and saves 30 minutes to an hour on each root-cause investigation. A manufacturer sets up agents for its own processes, and the same agent can answer an engineer's question, help an operator during a task, or run on its own on a schedule. Teams decide which actions an agent can take on its own, which need a person's approval, and which are off limits.
New at Operations Calling 2026, Tulip showed agents running inside automated workflows, where the agent handles the judgment and the workflow applies fixed rules. In the keynote demonstration, when an operator raised an Andon, a call for help on the line, an agent decided whether maintenance, inventory, or a supervisor should respond, and the workflow routed the call, notified that person, and recorded what happened. Agents in automated workflows are expected to be generally available by the end of 2026.
Factory Playback: see what happens on the line and act on it
Factory Playback, introduced last year with NVIDIA, puts video from cameras already on the floor on the same timeline as machine and order data. An engineer can click an event to see what happened on the floor next to the data from that moment, and follow how people, materials, robots, and automated guided vehicles moved through the space. New human activity detection shows where time goes at a workstation, whether a procedure was followed, and whether a part was assembled to spec. Factory Playback can also start work from what it sees. In the keynote demonstration on a bottling line, activity detection flagged that a line clearance may have started, a vision-language model checked what was happening, and an AI agent confirmed the clearance against the standard operating procedure and recorded it in Tulip. Video is processed at the plant, next to the cameras. Tulip developed Factory Playback's agent capabilities in close collaboration with NVIDIA.
"Factory Playback is about giving factories the ability to see, and giving the production system the ability to act on what it sees, for safety, for efficiency, and for quality. On a factory floor, an agent's actions reach machines and people, so it needs controls that keep it doing the right thing and points where a person stays in the loop," said Rony Kubat, Co-founder and CIO of Tulip.
Industrial Connectivity, Native MCP Connectors, Composable AI Agents, and Factory Playback make up the Run & Orchestrate announcements from Operations Calling 2026. To learn more, read the full Operations Calling 2026 announcement, explore Tulip AI, or book a demo.
About Tulip
Tulip is the Frontline Operations Platform. Manufacturers in regulated and complex industries use Tulip to build apps, agents, and automations in no-code, with AI, or in code, and to run them on one governed, composable platform with end-to-end traceability and the full context of the operation. A spinoff from MIT, Tulip is headquartered in Somerville, MA, with offices in Germany, Hungary, Singapore, Israel, and Japan.
This press release may include references to features that are still in development and not yet generally available. Customers should make purchase decisions based on features that are currently available.
Media contact: press@tulip.co