Interface McpServerConfig

Type Declaration

• OptionalhiddenPaths?: string[]

  Deprecated

  Use hideComponents instead. Dot-path strings of tools/domains to suppress.

• OptionalhideComponents?: { allModels?: boolean; domains?: string[]; paths?: readonly string[] }

  Fine-grained control over which tools are exposed to AI clients. All suppressed items are completely omitted — they will not appear in tool listings or START_HERE output.

  Dot-path format for paths:

  • 'myDomain' — hides the entire domain and all its features/models.
  • 'myDomain.myFeature' — hides one specific feature.
  • 'myDomain.cruds' — hides all model CRUD tools for a domain.
  • 'myDomain.cruds.MyModel' — hides CRUD tools for one specific model.

  • OptionalallModels?: boolean

    When true, suppresses ALL model CRUD tools across every domain.

  • Optionaldomains?: string[]

    Domain names to hide entirely (domain will not appear in list_domains).

  • Optionalpaths?: readonly string[]

    Dot-path strings identifying individual tools, features, or model CRUD groups to suppress.

• Optionallogging?: {
      requestLogGetData?: (req: Request) => Record<string, any>;
      requestLogLevel?: LogLevelNames;
      responseLogGetData?: (req: Request) => Record<string, any>;
      responseLogLevel?: LogLevelNames;
  }

  Logging configuration for inbound tool calls and their responses. All logging uses the Node-in-Layers logger already present in the context.

  • OptionalrequestLogGetData?: (req: Request) => Record<string, any>

    Optional callback that extracts additional structured fields to merge into the request log entry. Return a plain object of key/value pairs.

  • OptionalrequestLogLevel?: LogLevelNames

    Log level used when logging the incoming tool request. Defaults to 'info'. One of the LogLevelNames string literals from @node-in-layers/core (e.g. 'info', 'warn', 'error', 'debug').

  • OptionalresponseLogGetData?: (req: Request) => Record<string, any>

    Optional callback that extracts additional structured fields to merge into the response log entry. Return a plain object of key/value pairs.

  • OptionalresponseLogLevel?: LogLevelNames

    Log level used when logging the tool response. Defaults to 'info'. One of the LogLevelNames string literals from @node-in-layers/core (e.g. 'info', 'warn', 'error', 'debug').

• server: {
      connection:
          | Without<
              Readonly<{ type: "cli" }>,
              Readonly<
                  {
                      headers?: Readonly<Record<string, string>>;
                      retry?: Readonly<{ attempts: number; backoff: number }>;
                      timeout?: number;
                      type: "http";
                      url: string;
                  },
              >,
          > & Readonly<
              {
                  headers?: Readonly<Record<string, string>>;
                  retry?: Readonly<{ attempts: number; backoff: number }>;
                  timeout?: number;
                  type: "http";
                  url: string;
              },
          >
          | Without<
              Readonly<
                  {
                      headers?: Readonly<Record<string, string>>;
                      retry?: Readonly<{ attempts: number; backoff: number }>;
                      timeout?: number;
                      type: "http";
                      url: string;
                  },
              >,
              Readonly<{ type: "cli" }>,
          > & Readonly<{ type: "cli" }>;
  }

  Transport configuration. Must specify a connection that selects either CLI (stdio) or HTTP transport.

  • connection:
        | Without<
            Readonly<{ type: "cli" }>,
            Readonly<
                {
                    headers?: Readonly<Record<string, string>>;
                    retry?: Readonly<{ attempts: number; backoff: number }>;
                    timeout?: number;
                    type: "http";
                    url: string;
                },
            >,
        > & Readonly<
            {
                headers?: Readonly<Record<string, string>>;
                retry?: Readonly<{ attempts: number; backoff: number }>;
                timeout?: number;
                type: "http";
                url: string;
            },
        >
        | Without<
            Readonly<
                {
                    headers?: Readonly<Record<string, string>>;
                    retry?: Readonly<{ attempts: number; backoff: number }>;
                    timeout?: number;
                    type: "http";
                    url: string;
                },
            >,
            Readonly<{ type: "cli" }>,
        > & Readonly<{ type: "cli" }>

    Transport connection — either { type: 'cli' } or { type: 'http', url: '...', ... }.

    See

    Connection

• OptionalstartHere?: {
      description?: string;
      examplesOfUse?: readonly Readonly<
          {
              description?: string;
              details?: string;
              example?: string;
              name: string;
              tags?: string[];
              value?: string;
          },
      >[];
      hideDefaultSystemEntries?: boolean;
      includeDomains?: boolean;
      includeFeatures?: boolean;
      name?: string;
  }

  Configuration for the built-in START_HERE tool. This tool is always registered and is the recommended first call for any AI client — it returns a system overview that helps the client navigate the available domains, features, and models.

  • Optionaldescription?: string

    Override the tool description shown to the AI client.

  • OptionalexamplesOfUse?: readonly Readonly<
        {
            description?: string;
            details?: string;
            example?: string;
            name: string;
            tags?: string[];
            value?: string;
        },
    >[]

    Optional worked examples of end-to-end usage flows to include in the START_HERE response. Prefer JSON object examples with minimal prose.

    See

    SystemUseExample

  • OptionalhideDefaultSystemEntries?: boolean

    When true, suppresses the built-in Node-in-Layers navigation documentation from the response. Generally not recommended unless you are providing a completely custom system description.

  • OptionalincludeDomains?: boolean

    When true, the response includes the domain list (equivalent to calling list_domains first).

  • OptionalincludeFeatures?: boolean

    When true, the response includes the full feature list per domain (equivalent to calling list_features for every domain). Implies includeDomains: true.

  • Optionalname?: string

    Override the tool name. Default: 'START_HERE'.

• Optionalstateful?: boolean

  When true the HTTP transport manages persistent sessions (one McpServer instance per session). When false (the default) each HTTP request spins up a fresh server instance (stateless mode).

  Stateful mode is required when you need server-sent events or long-lived session state. Stateless mode is simpler and works well for most REST-style AI tool calls.

• OptionalsystemDescription?: { description?: string; version?: string }

  Static metadata injected into the START_HERE response.

  • Optionaldescription?: string

    Human-readable description of the system shown to the AI client.

  • Optionalversion?: string

    System version string shown to the AI client.

• Optionalversion?: string

  Semver string reported by the MCP server during the initialize handshake. Defaults to '1.0.0' if omitted.