FlowMCP Server

What is FlowMCP Server?

FlowMCP Server provides two server implementations for the FlowMCP framework: a LocalServer for local stdio-based execution and a RemoteServer for networked usage via HTTP and SSE. It is designed for developers building CLI tools, testing environments, frontend apps, and remote agents.

How to use FlowMCP Server?

Install the package (flowmcp-server), create a LocalServer or RemoteServer instance, prepare schema activations with FlowMCP.prepareActivations(...), add them via addActivationPayloads(...), then call start(). See the code examples in the README for local and remote configurations.

Key features of FlowMCP Server

• Lightweight local server using stdin/stdout with no extra dependencies
• Remote server supporting three transport protocols: SSE, stickyStreamable, statelessStreamable
• Optional Bearer token authentication for remote endpoints
• Configurable server metadata (name, description, version) via setConfig
• Compatible with FlowMCP framework and SchemaImporter for schema activation

Use cases of FlowMCP Server

• Run FlowMCP schemas locally in a CLI tool or development environment
• Expose schemas over HTTP for remote AI agents or web frontends
• Test schema activations with a simple local server before deploying
• Serve multiple routes with different transport protocols and authentication

FAQ from FlowMCP Server

What transport protocols does the RemoteServer support?

It supports sse (Server-Sent Events), stickyStreamable (HTTP with reusable sessions), and statelessStreamable (stateless POST-based HTTP).

How do I enable authentication on the RemoteServer?

Set the bearer option when calling addActivationPayloads(). Incoming requests must then include the header Authorization: Bearer <TOKEN>. Missing or invalid tokens return 401 or 403.

Can I customize the server name, description, or version?

Yes. Use localServer.setConfig({ overwrite: { serverDescription: { name, description, version } } }) or the equivalent on RemoteServer.

What version of FlowMCP Server is required?

FlowMCP Server version 1.2.0+ is required, and it works with FlowMCP Schema spec version 1.2.2 (as shown in the example).

Where does the data (schemas) come from?

Schemas are imported using the SchemaImporter package, which fetches them with SchemaImporter.get({ withSchema: true }). The README does not specify storage location.