About the Semaphore MCP Server
- Last Updated: November 20, 2026
- 4 minute read
- Semaphore
- Version 5.13
- Documentation
Starting with Semaphore 5.13.0, the Semaphore MCP Server enables external AI clients and custom AI agents to interact with Semaphore Knowledge Model Management (KMM) through the Model Context Protocol (MCP).
MCP provides a standardized way for AI applications to discover and use tools exposed by external systems. With the Semaphore MCP Server, you can retrieve information from knowledge models and perform supported KMM operations without building a separate integration for each operation.
The MCP Server is intended for developers, technical integrators, and organizations building AI-powered solutions that incorporate Semaphore knowledge models.
What can you do with the MCP Server?
You can use the Semaphore MCP Server to:
- Explore knowledge models. List available models and retrieve information about their structure.
- Find concepts. Search for concepts, inspect their properties, and retrieve their labels.
- Manage model content. Create and update concepts, labels, relationships, and metadata using supported MCP tools.
- Work with tasks. Perform supported task operations and commit task changes to the master model.
- Import RDF data. Import supported RDF triples into KMM.
For example, you can ask an AI assistant connected to Semaphore to list your available knowledge models, find concepts related to a particular topic, or retrieve the labels of a concept.
For practical examples, see Using MCP Tools: Examples.
How the MCP Server works
The Semaphore MCP Server acts as an interface between external MCP-compatible clients and Semaphore KMM.
A typical interaction follows these steps:
- Configure an MCP client with the Semaphore MCP Server endpoint and the authentication token required for your deployment environment.
- Connect the client to the server.
- The client discovers the available MCP tools.
- Ask the client to perform a task, such as finding a concept in a knowledge model.
- The client invokes the appropriate MCP tools.
- The MCP Server performs the supported KMM operations according to the permissions of the authenticated user.
- The client presents the results.
The operations available to a client depend on the tools exposed by the MCP Server and the permissions assigned to the authenticated user.
Supported functionality
The Semaphore MCP Server exposes more than 40 tools for working with KMM data.
These tools cover the following areas:
- Models and tasks: Retrieve and manage knowledge models and tasks, including supported task operations and committing task changes to the master model.
- Concepts: Retrieve, create, and modify concepts.
- Concept classes and concept schemes: Work with model structures and concept organization.
- Labels: Retrieve and manage concept labels.
- Relationships: Work with relationships between concepts.
- Metadata: Retrieve metadata type definitions and manage supported metadata values.
- Bulk import: Import supported RDF triples into KMM.
Note: Some concept retrieval tools return only a subset of concept properties. For example, getConceptAsJsonByUri does not return all metadata values associated with a concept.
For a complete reference, see Available MCP Tools.
MCP Server and AI Workflows
The MCP Server and AI Workflows are related but distinct Semaphore capabilities.
AI Workflows provide built-in AI-assisted functionality through the KMM user interface:
- Model from Spec creates knowledge models from user-provided specifications, prompts, and optional supporting documents.
- Enrichment extends existing knowledge models with AI-generated concepts, relationships, labels, and metadata.
The MCP Server provides tools that external clients and custom agents can use to perform supported KMM operations.
The built-in AI Workflows agents use MCP internally to perform KMM operations. However, the MCP Server does not expose the built-in Model from Spec or Enrichment agents as directly callable tools.
Developers can build their own AI agents using the available MCP tools, subject to supported operations and permissions.
For more information, see About AI Workflows.
Supported clients and connections
The Semaphore MCP Server uses HTTP transport.
For on-premises deployments, the MCP Server endpoint follows this format:
http://<semaphore_studio_url>/ai/mcp
Replace <semaphore_studio_url> with the hostname and port of your Semaphore Studio instance.
For Progress Data Cloud deployments, use the MCP Server endpoint configured for your cloud environment.
Authentication depends on the deployment environment:
- On-premises: A Semaphore access token passed in the
X-Api-KeyHTTP header. - Progress Data Cloud: A JWT token passed in the
AuthorizationHTTP header using the Bearer scheme.
Clients must support HTTP transport and custom authentication headers.
The MCP Server supports connections from compatible clients, including:
- Visual Studio Code with GitHub Copilot Chat: Configure the MCP Server in
.vscode/mcp.jsonand use Agent mode to access Semaphore tools. - GitHub Copilot CLI: Configure the MCP Server manually using the Copilot CLI MCP configuration file.
- Other MCP-compatible clients: Clients that support the required HTTP transport and authentication headers may also connect.
For connection requirements and setup instructions, see:
Limitations
The MCP Server provides access only to the KMM operations exposed by its available tools.
The following capabilities are not supported through MCP:
- Direct invocation of the built-in Model from Spec and Enrichment agents.
- OWL-specific functionality.
Some operations are subject to additional restrictions, including limitations related to model deletion, task management, and bulk RDF imports.
For more information, see Limitations and Troubleshooting.