MCP & AI Development
AI agents work within your ERP processes, inside the low-code layer's boundaries.
MCP (Model Context Protocol) (opens in new tab) connects AI agents directly to Nuclos and lets them act within your ERP processes: reading data, triggering workflows, writing back results.
The low-code layer defines the boundary. Agents have no direct access to the underlying data structures. This constraint is intentional: it keeps the solution space within validated platform patterns and makes AI-generated results maintainable and predictable.
Documentation status
In the Nuclos wiki, the “Nuclos AI & MCP” section is still marked as coming soon. Updated pages for Nuclos 5, Nuclosian, and MCP are not available there yet.
Until the wiki documentation catches up, Nuclos Workspace and the knowledge article on MCP for developers are the current entry points on this site.
Setup in four steps
-
Provision an instance
Install Nuclos 5 or provision it in the cloud. On-premise or Novabit Cloud.
-
Configure the MCP endpoint
Add the MCP endpoint to your AI agent or IDE, for example Claude or Cursor.
-
Authenticate with a token
Access via an API token with a defined role and permissions.
-
The agent works within its role
The agent reads, writes, and triggers actions, within the permissions of the token role.
Current MCP entry points
Workspace for practice, knowledge article for the technical overview.
More topics in the Dev Portal
Technical entry points for architecture, integration, and operations.
Ready for Nuclos Workspace?
Nuclos Workspace is the environment for modelling, configuration, and self-build with Nuclosian, MCP, and low-code.