Open Industrial lets MCP-compatible AI tools propose governed workflows

By AI, Created 5:09 PM UTC, May 18, 2026, /AGP/ – Open Industrial has expanded its MCP server so external AI tools can propose data workflows into the platform, while human approval remains required before deployment. The company also added an API Source node that turns REST endpoints into live data feeds without custom code.

Why it matters: - Open Industrial is trying to remove a common tradeoff for industrial and regulated teams: use the AI tools developers prefer without losing governance controls. - The update lets teams keep a human approval gate, audit trail and attribution metadata while still using external AI agents to draft workflow proposals. - The API Source node could shorten the time it takes to bring REST endpoints into governed workflows.

What happened: - Open Industrial, a governed data-to-AI platform by Fathym, announced that any external MCP-compatible AI tool can now propose data workflows directly into the platform. - The company said nothing deploys without human review and approval. - Open Industrial also introduced an API Source node for connecting REST endpoints as live data feeds. - Both capabilities are available immediately at Open Industrial.

The details: - The expanded MCP server accepts proposals from tools including Claude Code, GitHub Copilot, custom agents and other frontier models accessible through MCP. - A developer can connect an AI tool to an Open Industrial workspace, describe a workflow or integration in natural language, and receive a structured proposal inside the platform’s review interface. - Every proposal follows the same governance workflow and is logged with the same attribution metadata. - The bidirectional MCP setup lets a developer identify needed connections, generate a flow proposal, approve it with a single click and then query across the provisioned sources without leaving the coding environment. - The API Source node connects any REST endpoint as a live data source. - Polling intervals, authentication and response mapping are configured through the workspace’s visual inspector, with no code required. - A manual Pull Now mode lets teams validate a new endpoint before scheduling deployment. - Endpoint health is monitored continuously, and failures surface immediately before clearing automatically when the endpoint recovers.

Between the lines: - Open Industrial is positioning MCP as the bridge between AI-assisted development and enterprise governance. - The pitch is not that AI deploys autonomously. The pitch is that AI can draft the work, while humans still approve the final action. - That framework may appeal to industrial and life sciences teams that need faster integration work but cannot relax controls.

What’s next: - Open Industrial will likely use the new MCP workflow proposal feature and API Source node to push more external AI tools into governed enterprise environments. - The company said the capabilities are now live, so teams can start testing the workflow proposal and REST-source features immediately.

The bottom line: - Open Industrial is betting that governed AI adoption will scale faster when developers can stay in their preferred tools and still route every action through human approval.