Sidearm MCP Server

Instructions

Implementation Reference

• src/tools/detect-membership.ts:5-65 (handler)

  Main implementation of the detect_membership tool. Contains the server.tool registration and the async handler function that executes the membership inference logic by calling the API endpoint /api/v1/detect/membership with the provided parameters.

  export function register(server: McpServer, api: ApiClient): void {
    server.tool(
      "detect_membership",
      "Run membership inference to determine whether your protected content was used to train " +
        "a suspect AI model. Provide content IDs (from your registered media) and the model to test. " +
        "Methods: pattern (watermark detection), statistical (distribution analysis), " +
        "combined (both). Returns a job_id — use check_job to poll for results.",
      {
        content_ids: z
          .array(z.string())
          .min(1)
          .describe("UUIDs of your registered media to test against the suspect model"),
        suspect_model: z
          .string()
          .describe("Identifier or name of the AI model suspected of training on your content"),
        method: z
          .enum(["pattern", "statistical", "combined"])
          .optional()
          .describe("Inference method. Default: combined"),
        tags: z
          .array(z.string())
          .optional()
          .describe("Tags for organizing and filtering"),
      },
      async (params) => {
        try {
          const body: Record<string, unknown> = {
            content_ids: params.content_ids,
            suspect_model: params.suspect_model,
          };
          if (params.method) body.method = params.method;
          if (params.tags) body.tags = params.tags;
  
          const result = await api.post("/api/v1/detect/membership", body);
          const res = result as { job_id: string };
  
          return {
            content: [
              {
                type: "text" as const,
                text:
                  `Membership inference job created.\n\n` +
                  `Job ID: ${res.job_id}\n\n` +
                  `Use check_job with this job_id to poll for results.`,
              },
            ],
          };
        } catch (err) {
          return {
            content: [
              {
                type: "text" as const,
                text: `Error: ${err instanceof Error ? err.message : String(err)}`,
              },
            ],
            isError: true as const,
          };
        }
      },
    );
  }

• src/tools/detect-membership.ts:12-28 (schema)

  Zod schema definition for the detect_membership tool input parameters, including content_ids (array of UUIDs), suspect_model (string identifier), method (enum: pattern, statistical, combined, optional), and tags (array of strings, optional).

  {
    content_ids: z
      .array(z.string())
      .min(1)
      .describe("UUIDs of your registered media to test against the suspect model"),
    suspect_model: z
      .string()
      .describe("Identifier or name of the AI model suspected of training on your content"),
    method: z
      .enum(["pattern", "statistical", "combined"])
      .optional()
      .describe("Inference method. Default: combined"),
    tags: z
      .array(z.string())
      .optional()
      .describe("Tags for organizing and filtering"),
  },

• src/tools/detect-membership.ts:29-63 (handler)

  Async handler function that processes the detect_membership request. Validates parameters, constructs the request body, calls the API client's post method to /api/v1/detect/membership, and returns formatted results with the job_id or error message.

  async (params) => {
    try {
      const body: Record<string, unknown> = {
        content_ids: params.content_ids,
        suspect_model: params.suspect_model,
      };
      if (params.method) body.method = params.method;
      if (params.tags) body.tags = params.tags;
  
      const result = await api.post("/api/v1/detect/membership", body);
      const res = result as { job_id: string };
  
      return {
        content: [
          {
            type: "text" as const,
            text:
              `Membership inference job created.\n\n` +
              `Job ID: ${res.job_id}\n\n` +
              `Use check_job with this job_id to poll for results.`,
          },
        ],
      };
    } catch (err) {
      return {
        content: [
          {
            type: "text" as const,
            text: `Error: ${err instanceof Error ? err.message : String(err)}`,
          },
        ],
        isError: true as const,
      };
    }
  },

• src/index.ts:14-14 (registration)

  Import statement for the detect_membership register function from the tools module.

  import { register as detectMembership } from "./tools/detect-membership.js";

• src/index.ts:56-56 (registration)

  Registration call that registers the detect_membership tool with the MCP server instance.

  detectMembership(server, api);