What carbon does
The carbon MCP server gives AI agents access to operations in Carbon, an open-source manufacturing platform combining enterprise resource planning, material requirements planning, manufacturing execution and quality management. Carbon stores these areas in one Postgres schema, so engineering, planning, production, quality and accounting work from shared records.
The server is intended for manufacturing workflows such as quoting, sales orders, purchasing, inventory, invoicing, job costing, material planning, production execution, quality management and lot or serial traceability. It also exposes capabilities related to bills of materials, routings, revisions, configuration, workflows, maintenance and assets. The available operation set follows Carbon’s API rather than presenting a separate data model.
This listing belongs to the Industrial & IoT category and is relevant to teams building agent workflows around production data, shop-floor execution or manufacturing administration.
How it works
Carbon provides the same manufacturing operations through HTTP and MCP interfaces. MCP presents those operations as tools for hosted or local AI agents, while HTTP exposes them through POST requests under /api/v1/{module}/{operation}. The API has a published OpenAPI specification for typed client generation, and both interfaces use the same arguments.
Each operation validates its input, updates dependent records and checks the permissions of the authenticated identity. Access is therefore tied to the API key or signed-in user used by the connection. Carbon also applies tenant-scoped records, row-level security and role- or attribute-based access controls across its application, API and MCP server.
The carbon MCP server can be used with Carbon Cloud or with a self-hosted deployment. Self-hosted installations expose the same API beneath /api/v1, allowing the deployment to run in a company VPC, on-premises environment or air-gapped network.
Setup and configuration
Create an API key in Carbon under Settings → API Keys, then authenticate requests with the key. The documented HTTP example uses the CARBON_API_KEY environment variable as a bearer token. MCP connections are scoped to the permissions of that key or to the signed-in user, so the selected identity determines which records and operations an agent can use.
Carbon is available as a managed cloud application or as a self-hosted installation. Cloud provides the managed application, database, updates and backups.
Tools and capabilities
The carbon MCP server provides permission-aware access to Carbon API operations across areas including:
- Quotes, sales orders, purchasing, inventory, invoicing and actual job costs
- Demand and supply planning, versioned bills of materials, routings and finite-capacity scheduling
- Digital travelers, operator terminals, 3D work instructions, barcode capture and labor capture
- Inspections, first article inspection, nonconformance, corrective and preventive action, calibration and risk management
- Forward and backward lot and serial genealogy
- Engineering revisions, change orders, supersession and product configuration
- Rule-based workflows, scheduled maintenance, fixed assets and kanban replenishment
The surrounding Carbon platform also lists integrations with Onshape, SolidWorks, Paperless Parts, Linear, Jira, Slack, Ramp, Stripe and Zebra.
Limitations and notes
The carbon MCP server is permission-aware, so an agent cannot be assumed to have unrestricted access to Carbon data. Its available actions depend on the API key or signed-in user and the deployment’s access controls.
Review Carbon’s MCP documentation before connecting a particular agent client. Carbon is licensed under AGPL-3.0.