Port MCP Server – Developer Portal Access
Port provides a remote Model Context Protocol server for interacting with a Port.io developer portal from AI-enabled tools such as Cursor, VS Code, Claude, and Codex CLI. It gives agents access to catalog data, scorecards, actions, workflows, and permission-aware developer portal operations, subject to the connected user's Port permissions and configured MCP headers.
Overview
Port is an internal developer portal and software catalog platform. The Port MCP Server is Port's remote Model Context Protocol interface for bringing developer portal context into AI tools. It lets an AI agent work with the same Port organization that the user authorizes, so answers and actions are grounded in live catalog, ownership, scorecard, and workflow data rather than static documentation.
What the MCP server enables
Port documents the MCP server as a bridge between LLM-powered tools and a Port.io developer portal. Depending on the user's role and permissions, the server can expose tools for developer and builder workflows. Verified tool areas include listing and querying blueprints and entities, describing the connected user, listing actions and workflows, checking action permissions, running permitted actions, triggering workflows, and tracking action or workflow execution status.
The available tool set is permission-aware. Developers typically use it to query services, owners, dependencies, health, scorecards, and approved self-service workflows. Builders can use it to help configure Port data models and portal structures such as blueprints, scorecards, and actions where their permissions allow.
When to use it
Use Port MCP when an AI assistant needs live developer portal context: finding service owners, identifying production services that fail scorecards, checking dependencies, exploring catalog data, or helping a developer understand what needs to be done to onboard or improve a service. It is also useful for governed operational workflows, such as running approved self-service actions or triggering workflows from an IDE or chat environment.
Connection and authentication
Port provides regional remote MCP URLs. EU organizations use https://mcp.port.io/v1; US organizations use https://mcp.us.port.io/v1. Interactive clients authenticate with Port through an OAuth flow in the browser. For non-interactive automation, Port also documents token-based authentication using Port client credentials to obtain an access token. Some clients can connect directly to the remote URL, while others use the documented mcp-remote command wrapper.
Key considerations
Port MCP access is governed by the connected Port organization, the user's permissions, and optional headers. The x-read-only-mode header can be set to 1 to hide write tools and expose only read-only tools. The x-allowed-actions-to-run header can restrict which action identifiers are available through the action-running tool. If an organization uses SSO, Port notes that the SSO connection may need domain-level authentication enabled for the MCP OAuth flow to work correctly.
Supported Transports
streamable_http
URL: https://mcp.port.io/v1
streamable_http
URL: https://mcp.us.port.io/v1
stdio
Command: npx
Args:
-ymcp-remotehttps://mcp.port.io/v1
Frequently Asked Questions
- When should an AI agent use the Port MCP Server?
- Use it when the agent needs live Port developer portal context, such as querying the software catalog, finding service owners, checking scorecard status, reviewing dependencies, or helping a developer run governed self-service workflows from an IDE or chat tool.
- What does the Port MCP Server add to an AI agent's capabilities?
- It adds permission-aware access to Port data and operations. The agent can use documented tools for catalog entities and blueprints, user context, actions, workflows, scorecards, and execution tracking instead of relying on static model knowledge.
- What can the AI agent access or manage through Port MCP?
- Depending on the connected user's role and permissions, Port MCP can list and query blueprints and entities, inspect user details, list actions and workflows, check action permissions, run permitted actions, trigger workflow runs, and track action or workflow status.
- How is authentication configured for the Port MCP Server?
- Interactive MCP clients use Port's browser-based OAuth flow against the selected Port organization. Port also documents token-based authentication for CI/CD or other non-interactive environments using Port client credentials to obtain an access token.
- Which transport should be used for the Port MCP Server?
- Use the regional remote HTTP MCP URL directly when the client supports remote MCP servers: `https://mcp.port.io/v1` for EU or `https://mcp.us.port.io/v1` for US. Use the documented `npx mcp-remote` stdio wrapper for clients or environments that require a command-based MCP configuration.