@yawlabs/mcp-compliance 0.6.0 → 0.7.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/index.js CHANGED
@@ -3,7 +3,7 @@
3
3
  // src/index.ts
4
4
  import { createRequire as createRequire3 } from "module";
5
5
  import chalk2 from "chalk";
6
- import { Command } from "commander";
6
+ import { Command, Option } from "commander";
7
7
 
8
8
  // src/mcp/server.ts
9
9
  import { createRequire as createRequire2 } from "module";
@@ -71,7 +71,7 @@ function computeScore(tests) {
71
71
 
72
72
  // src/types.ts
73
73
  var TEST_DEFINITIONS = [
74
- // ── Transport (10 tests) ─────────────────────────────────────────
74
+ // ── Transport (13 tests) ─────────────────────────────────────────
75
75
  {
76
76
  id: "transport-post",
77
77
  name: "HTTP POST accepted",
@@ -96,8 +96,8 @@ var TEST_DEFINITIONS = [
96
96
  category: "transport",
97
97
  required: false,
98
98
  specRef: "basic/transports#streamable-http",
99
- description: "Verifies that sending a JSON-RPC notification (no id field) returns HTTP 202 Accepted with no body. Per spec, servers MUST return 202 for notifications.",
100
- recommendation: "Detect JSON-RPC messages without an id field and return HTTP 202 with an empty body. Do not attempt to send a JSON-RPC response for notifications."
99
+ description: "Verifies that sending a JSON-RPC notification (no id field) returns exactly HTTP 202 Accepted with no body. Per spec, servers MUST return 202 \u2014 not 200 or 204.",
100
+ recommendation: "Detect JSON-RPC messages without an id field and return HTTP 202 with an empty body. Do not return 200 or 204 \u2014 the spec requires exactly 202 Accepted."
101
101
  },
102
102
  {
103
103
  id: "transport-session-id",
@@ -108,6 +108,24 @@ var TEST_DEFINITIONS = [
108
108
  description: "Tests that the server returns HTTP 400 when MCP-Session-Id header is missing on requests after initialization (when the server issued a session ID).",
109
109
  recommendation: "If your server issues an MCP-Session-Id header in the initialize response, reject subsequent requests that omit this header with HTTP 400."
110
110
  },
111
+ {
112
+ id: "transport-session-invalid",
113
+ name: "Returns 404 for unknown session ID",
114
+ category: "transport",
115
+ required: false,
116
+ specRef: "basic/transports#streamable-http",
117
+ description: "Sends a request with a fabricated MCP-Session-Id and verifies the server returns HTTP 404. Per spec, servers managing sessions MUST return 404 for unrecognized session IDs.",
118
+ recommendation: "Return HTTP 404 (Not Found) for requests with an MCP-Session-Id that does not match any active session. Do not return 400 \u2014 that is for missing session IDs."
119
+ },
120
+ {
121
+ id: "transport-content-type-reject",
122
+ name: "Rejects non-JSON request Content-Type",
123
+ category: "transport",
124
+ required: false,
125
+ specRef: "basic/transports#streamable-http",
126
+ description: "Sends a POST with Content-Type: text/plain instead of application/json and verifies the server rejects it with a 4xx status.",
127
+ recommendation: "Validate the Content-Type header on incoming POST requests. Reject requests that are not application/json with HTTP 415 (Unsupported Media Type) or 400."
128
+ },
111
129
  {
112
130
  id: "transport-get",
113
131
  name: "GET returns SSE stream or 405",
@@ -162,7 +180,16 @@ var TEST_DEFINITIONS = [
162
180
  description: "Sends multiple JSON-RPC requests in parallel and verifies the server responds to all with correct matching IDs. Tests that the server can handle concurrent connections.",
163
181
  recommendation: "Ensure your server can handle multiple simultaneous requests. Each response must include the correct id matching the request. Use async handlers or connection pooling."
164
182
  },
165
- // ── Lifecycle (12 tests) ─────────────────────────────────────────
183
+ {
184
+ id: "transport-sse-event-field",
185
+ name: "SSE responses include event: message",
186
+ category: "transport",
187
+ required: false,
188
+ specRef: "basic/transports#streamable-http",
189
+ description: "Sends a request with Accept: text/event-stream and checks that SSE responses include the event: message field. Per spec, servers MUST set event: message for JSON-RPC messages in SSE streams.",
190
+ recommendation: 'Include "event: message" before each "data:" line in your SSE responses. This is required by the MCP spec for JSON-RPC messages sent over SSE.'
191
+ },
192
+ // ── Lifecycle (15 tests) ─────────────────────────────────────────
166
193
  {
167
194
  id: "lifecycle-init",
168
195
  name: "Initialize handshake",
@@ -235,6 +262,33 @@ var TEST_DEFINITIONS = [
235
262
  description: "Verifies that the JSON-RPC response id matches the request id sent by the client. This is a fundamental JSON-RPC 2.0 requirement.",
236
263
  recommendation: "Copy the id field from the request into the response. This is a core JSON-RPC 2.0 requirement. Check that your framework does not modify or discard the request ID."
237
264
  },
265
+ {
266
+ id: "lifecycle-string-id",
267
+ name: "Supports string request IDs",
268
+ category: "lifecycle",
269
+ required: false,
270
+ specRef: "basic",
271
+ description: "Sends a request with a string id instead of a number. JSON-RPC 2.0 allows both string and number IDs. The server must echo back the exact string id in the response.",
272
+ recommendation: "Ensure your JSON-RPC implementation supports both string and number request IDs. Echo the id back exactly as received, preserving its type."
273
+ },
274
+ {
275
+ id: "lifecycle-version-negotiate",
276
+ name: "Handles unknown protocol version",
277
+ category: "lifecycle",
278
+ required: false,
279
+ specRef: "basic/lifecycle#version-negotiation",
280
+ description: 'Sends an initialize request with a future protocol version ("2099-01-01") and verifies the server either negotiates down to a version it supports or returns an error.',
281
+ recommendation: "When the client requests an unsupported protocol version, respond with the closest version your server supports. Do not blindly accept unknown versions."
282
+ },
283
+ {
284
+ id: "lifecycle-reinit-reject",
285
+ name: "Rejects second initialize request",
286
+ category: "lifecycle",
287
+ required: false,
288
+ specRef: "basic/lifecycle#initialization",
289
+ description: "Sends a second initialize request within the same session. Per spec, the client MUST NOT send initialize more than once. The server should reject it.",
290
+ recommendation: "Track initialization state per session. Reject duplicate initialize requests with a JSON-RPC error or HTTP 4xx. Do not reset session state on re-initialization."
291
+ },
238
292
  {
239
293
  id: "lifecycle-logging",
240
294
  name: "logging/setLevel accepted",
@@ -264,12 +318,12 @@ var TEST_DEFINITIONS = [
264
318
  },
265
319
  {
266
320
  id: "lifecycle-progress",
267
- name: "Accepts progress notifications",
321
+ name: "Handles progress notifications gracefully",
268
322
  category: "lifecycle",
269
323
  required: false,
270
324
  specRef: "basic/utilities#progress",
271
- description: "Tests that the server accepts notifications/progress without error. Servers should handle progress notifications for request tracking.",
272
- recommendation: "Accept notifications/progress with progressToken, progress, and optional total fields. Ignore notifications for unknown progress tokens."
325
+ description: "Sends a notifications/progress to the server and verifies it does not error. Note: per spec, progress flows from server to client during long-running requests. This test validates the server handles unexpected notifications gracefully.",
326
+ recommendation: "Accept unknown notifications without returning an error. The server should not crash or return a non-2xx status for notifications it does not recognize."
273
327
  },
274
328
  // ── Tools (4 tests) ──────────────────────────────────────────────
275
329
  {
@@ -278,7 +332,7 @@ var TEST_DEFINITIONS = [
278
332
  category: "tools",
279
333
  required: false,
280
334
  specRef: "server/tools#listing-tools",
281
- description: "Calls tools/list and validates it returns an array of tool definitions. Required if the server declares tools capability.",
335
+ description: "Calls tools/list and validates it returns an array of tool definitions. Dynamically required at runtime if the server declares tools capability.",
282
336
  recommendation: "Implement the tools/list handler to return { tools: [...] } with an array of tool definition objects. Each tool needs at least a name and inputSchema."
283
337
  },
284
338
  {
@@ -315,7 +369,7 @@ var TEST_DEFINITIONS = [
315
369
  category: "resources",
316
370
  required: false,
317
371
  specRef: "server/resources#listing-resources",
318
- description: "Calls resources/list and validates it returns an array. Required if the server declares resources capability.",
372
+ description: "Calls resources/list and validates it returns an array. Dynamically required at runtime if the server declares resources capability.",
319
373
  recommendation: "Implement resources/list to return { resources: [...] } with an array of resource objects. Each resource needs at least a uri and name."
320
374
  },
321
375
  {
@@ -361,7 +415,7 @@ var TEST_DEFINITIONS = [
361
415
  category: "prompts",
362
416
  required: false,
363
417
  specRef: "server/prompts#listing-prompts",
364
- description: "Calls prompts/list and validates it returns an array. Required if the server declares prompts capability.",
418
+ description: "Calls prompts/list and validates it returns an array. Dynamically required at runtime if the server declares prompts capability.",
365
419
  recommendation: "Implement prompts/list to return { prompts: [...] } with an array of prompt objects. Each prompt needs at least a name field."
366
420
  },
367
421
  {
@@ -382,7 +436,7 @@ var TEST_DEFINITIONS = [
382
436
  description: "Tests cursor-based pagination on prompts/list. Validates nextCursor is a string if present and that fetching the next page works.",
383
437
  recommendation: "If you return nextCursor in prompts/list, ensure it is a string and that passing it back as cursor in the next request returns valid results."
384
438
  },
385
- // ── Error Handling (8 tests) ─────────────────────────────────────
439
+ // ── Error Handling (10 tests) ────────────────────────────────────
386
440
  {
387
441
  id: "error-unknown-method",
388
442
  name: "Returns JSON-RPC error for unknown method",
@@ -455,6 +509,24 @@ var TEST_DEFINITIONS = [
455
509
  description: "Calls tools/call with a nonexistent tool name and verifies the server returns an error response.",
456
510
  recommendation: "Return a JSON-RPC error or set isError: true when tools/call receives an unrecognized tool name. Do not return an empty success response."
457
511
  },
512
+ {
513
+ id: "error-capability-gated",
514
+ name: "Rejects methods for undeclared capabilities",
515
+ category: "errors",
516
+ required: false,
517
+ specRef: "basic/lifecycle#capability-negotiation",
518
+ description: "Calls list methods (tools/list, resources/list, prompts/list) for capabilities the server did NOT declare, and verifies the server returns an error instead of success.",
519
+ recommendation: "Return a JSON-RPC error (e.g., -32601 Method not found) for methods associated with capabilities not declared in your initialize response."
520
+ },
521
+ {
522
+ id: "error-invalid-cursor",
523
+ name: "Handles invalid pagination cursor gracefully",
524
+ category: "errors",
525
+ required: false,
526
+ specRef: "basic",
527
+ description: "Sends a garbage pagination cursor to a list method and verifies the server handles it gracefully \u2014 either returning an error or ignoring the invalid cursor.",
528
+ recommendation: "Validate pagination cursors before use. Return a JSON-RPC error for unrecognized cursors, or treat invalid cursors as a request for the first page."
529
+ },
458
530
  // ── Schema Validation (6 tests) ──────────────────────────────────
459
531
  {
460
532
  id: "tools-schema",
@@ -471,8 +543,8 @@ var TEST_DEFINITIONS = [
471
543
  category: "schema",
472
544
  required: false,
473
545
  specRef: "server/tools#annotations",
474
- description: "Validates tool annotation fields if present: readOnlyHint, destructiveHint, idempotentHint, openWorldHint should be booleans; title should be a string.",
475
- recommendation: "If you include annotations on tools, ensure readOnlyHint, destructiveHint, idempotentHint, and openWorldHint are booleans. Title must be a string."
546
+ description: "Validates tool annotation fields if present: readOnlyHint, destructiveHint, idempotentHint, openWorldHint should be booleans.",
547
+ recommendation: "If you include annotations on tools, ensure readOnlyHint, destructiveHint, idempotentHint, and openWorldHint are booleans. Note: title belongs on the Tool object, not inside annotations."
476
548
  },
477
549
  {
478
550
  id: "tools-title-field",
@@ -510,7 +582,7 @@ var TEST_DEFINITIONS = [
510
582
  description: "Validates every resource has a valid URI (parseable as a URL) and a name field.",
511
583
  recommendation: "Ensure every resource has a valid, parseable URI and a name field. Add description and mimeType for better client integration."
512
584
  },
513
- // ── Security: Auth & Transport (8 tests) ─────────────────────────
585
+ // ── Security: Auth & Transport (9 tests) ─────────────────────────
514
586
  {
515
587
  id: "security-auth-required",
516
588
  name: "Rejects unauthenticated requests",
@@ -558,12 +630,12 @@ var TEST_DEFINITIONS = [
558
630
  },
559
631
  {
560
632
  id: "security-oauth-metadata",
561
- name: "OAuth metadata endpoint exists",
633
+ name: "Protected Resource Metadata endpoint exists",
562
634
  category: "security",
563
635
  required: false,
564
636
  specRef: "basic/authorization",
565
- description: "Checks for a well-known OAuth authorization server metadata endpoint at /.well-known/oauth-authorization-server. If the server requires auth, it should advertise how to obtain tokens.",
566
- recommendation: "Publish an OAuth 2.0 Authorization Server Metadata document at /.well-known/oauth-authorization-server on your server's origin. Include issuer, token_endpoint, and supported grant types."
637
+ description: "Checks for a Protected Resource Metadata (RFC 9728) endpoint at /.well-known/oauth-protected-resource. Per MCP 2025-11-25, the MCP server publishes PRM with a resource identifier and authorization_servers array. Falls back to legacy /.well-known/oauth-authorization-server with a warning.",
638
+ recommendation: "Publish a Protected Resource Metadata document at /.well-known/oauth-protected-resource on your server's origin. Include 'resource' (your server's URL) and 'authorization_servers' (array of OAuth AS URLs). See RFC 9728."
567
639
  },
568
640
  {
569
641
  id: "security-token-in-uri",
@@ -583,6 +655,15 @@ var TEST_DEFINITIONS = [
583
655
  description: "If the server returns CORS headers, verifies that Access-Control-Allow-Origin is not set to wildcard (*). Wildcard CORS on an authenticated API allows cross-origin credential theft.",
584
656
  recommendation: 'Set Access-Control-Allow-Origin to specific trusted origins, not "*". If CORS is not needed (server-to-server only), do not send CORS headers at all.'
585
657
  },
658
+ {
659
+ id: "security-origin-validation",
660
+ name: "Validates Origin header on requests",
661
+ category: "security",
662
+ required: false,
663
+ specRef: "basic/transports#streamable-http",
664
+ description: "Sends a request with a suspicious Origin header (https://evil-rebinding-attack.example.com) and verifies the server rejects it. Per spec, servers MUST validate the Origin header to prevent DNS rebinding attacks.",
665
+ recommendation: "Validate the Origin header on all incoming requests. Reject requests from untrusted origins with HTTP 403. Maintain an allowlist of permitted origins."
666
+ },
586
667
  // ── Security: Input Validation (6 tests) ─────────────────────────
587
668
  {
588
669
  id: "security-command-injection",
@@ -768,7 +849,7 @@ function createIdCounter(start = 0) {
768
849
  }
769
850
  function parseSSEResponse(text) {
770
851
  const lines = text.split("\n");
771
- let lastJsonRpcResponse = null;
852
+ let firstJsonRpcResponse = null;
772
853
  let currentData = [];
773
854
  function flushEvent() {
774
855
  if (currentData.length === 0) return;
@@ -777,8 +858,8 @@ function parseSSEResponse(text) {
777
858
  if (!data.trim()) return;
778
859
  try {
779
860
  const parsed = JSON.parse(data);
780
- if (parsed.jsonrpc === "2.0" && parsed.id !== void 0) {
781
- lastJsonRpcResponse = parsed;
861
+ if (!firstJsonRpcResponse && parsed.jsonrpc === "2.0" && parsed.id !== void 0) {
862
+ firstJsonRpcResponse = parsed;
782
863
  }
783
864
  } catch {
784
865
  }
@@ -792,7 +873,7 @@ function parseSSEResponse(text) {
792
873
  }
793
874
  }
794
875
  flushEvent();
795
- return lastJsonRpcResponse;
876
+ return firstJsonRpcResponse;
796
877
  }
797
878
  async function mcpRequest(backendUrl, method, params, nextId, extraHeaders, timeout) {
798
879
  const id = nextId();
@@ -1002,6 +1083,32 @@ async function runComplianceSuite(url, options = {}) {
1002
1083
  return { passed: valid, details: `Content-Type: ${ct}` };
1003
1084
  }
1004
1085
  );
1086
+ await test(
1087
+ "transport-content-type-reject",
1088
+ "Rejects non-JSON request Content-Type",
1089
+ "transport",
1090
+ false,
1091
+ "basic/transports#streamable-http",
1092
+ async () => {
1093
+ const res = await request(backendUrl, {
1094
+ method: "POST",
1095
+ headers: { "Content-Type": "text/plain", Accept: "application/json, text/event-stream", ...userHeaders },
1096
+ body: JSON.stringify({ jsonrpc: "2.0", id: 99905, method: "ping" }),
1097
+ signal: AbortSignal.timeout(timeout)
1098
+ });
1099
+ await res.body.text();
1100
+ if (res.statusCode >= 400 && res.statusCode < 500) {
1101
+ return { passed: true, details: `HTTP ${res.statusCode} (incorrect Content-Type rejected)` };
1102
+ }
1103
+ if (res.statusCode >= 200 && res.statusCode < 300) {
1104
+ return {
1105
+ passed: false,
1106
+ details: `HTTP ${res.statusCode} \u2014 server accepted text/plain Content-Type (should require application/json)`
1107
+ };
1108
+ }
1109
+ return { passed: false, details: `HTTP ${res.statusCode}` };
1110
+ }
1111
+ );
1005
1112
  await test(
1006
1113
  "transport-get",
1007
1114
  "GET returns SSE stream or 405",
@@ -1016,7 +1123,8 @@ async function runComplianceSuite(url, options = {}) {
1016
1123
  signal: AbortSignal.timeout(timeout)
1017
1124
  });
1018
1125
  const body = await res.body.text();
1019
- const ct = (res.headers["content-type"] || "").toLowerCase();
1126
+ const rawCt = res.headers["content-type"];
1127
+ const ct = (Array.isArray(rawCt) ? rawCt[0] : rawCt || "").toLowerCase();
1020
1128
  if (res.statusCode === 405) {
1021
1129
  return { passed: true, details: "HTTP 405 Method Not Allowed (acceptable)" };
1022
1130
  }
@@ -1124,6 +1232,50 @@ async function runComplianceSuite(url, options = {}) {
1124
1232
  return { passed: valid, details: `Version: ${version3}` };
1125
1233
  }
1126
1234
  );
1235
+ await test(
1236
+ "lifecycle-version-negotiate",
1237
+ "Handles unknown protocol version",
1238
+ "lifecycle",
1239
+ false,
1240
+ "basic/lifecycle#version-negotiation",
1241
+ async () => {
1242
+ try {
1243
+ const futureRes = await mcpRequest(
1244
+ backendUrl,
1245
+ "initialize",
1246
+ {
1247
+ protocolVersion: "2099-01-01",
1248
+ capabilities: {},
1249
+ clientInfo: { name: "mcp-compliance", version: TOOL_VERSION }
1250
+ },
1251
+ createIdCounter(99960),
1252
+ userHeaders,
1253
+ timeout
1254
+ );
1255
+ const result = futureRes.body?.result;
1256
+ const error = futureRes.body?.error;
1257
+ if (error) {
1258
+ return {
1259
+ passed: true,
1260
+ details: `Server rejected unknown version with error: ${error.code} \u2014 ${error.message}`
1261
+ };
1262
+ }
1263
+ if (result?.protocolVersion) {
1264
+ const offered = result.protocolVersion;
1265
+ if (offered === "2099-01-01") {
1266
+ return {
1267
+ passed: false,
1268
+ details: 'Server accepted impossible future version "2099-01-01" \u2014 should offer a version it actually supports'
1269
+ };
1270
+ }
1271
+ return { passed: true, details: `Server negotiated down to ${offered} (correct)` };
1272
+ }
1273
+ return { passed: false, details: "No protocolVersion or error in response" };
1274
+ } catch {
1275
+ return { passed: true, details: "Connection rejected for unknown version (acceptable)" };
1276
+ }
1277
+ }
1278
+ );
1127
1279
  await test(
1128
1280
  "lifecycle-server-info",
1129
1281
  "Includes serverInfo",
@@ -1195,6 +1347,73 @@ async function runComplianceSuite(url, options = {}) {
1195
1347
  details: match ? `Request id=${res.requestId}, response id=${body.id} (match)` : `Request id=${res.requestId}, response id=${body.id} (MISMATCH)`
1196
1348
  };
1197
1349
  });
1350
+ await test("lifecycle-string-id", "Supports string request IDs", "lifecycle", false, "basic", async () => {
1351
+ const stringId = "compliance-test-string-id";
1352
+ const body = JSON.stringify({ jsonrpc: "2.0", id: stringId, method: "ping", params: {} });
1353
+ const res = await request(backendUrl, {
1354
+ method: "POST",
1355
+ headers: {
1356
+ "Content-Type": "application/json",
1357
+ Accept: "application/json, text/event-stream",
1358
+ ...buildHeaders()
1359
+ },
1360
+ body,
1361
+ signal: AbortSignal.timeout(timeout)
1362
+ });
1363
+ const text = await res.body.text();
1364
+ const rawCtStr = res.headers["content-type"];
1365
+ const ct = (Array.isArray(rawCtStr) ? rawCtStr[0] : rawCtStr || "").toLowerCase();
1366
+ let parsed;
1367
+ if (ct.includes("text/event-stream")) {
1368
+ parsed = parseSSEResponse(text);
1369
+ }
1370
+ if (!parsed) {
1371
+ try {
1372
+ parsed = JSON.parse(text);
1373
+ } catch {
1374
+ }
1375
+ }
1376
+ if (!parsed) return { passed: false, details: "Could not parse response" };
1377
+ if (parsed.id === stringId) {
1378
+ return { passed: true, details: `String id="${stringId}" echoed back correctly` };
1379
+ }
1380
+ if (parsed.id === void 0) {
1381
+ return { passed: false, details: "No id in response" };
1382
+ }
1383
+ return {
1384
+ passed: false,
1385
+ details: `String id="${stringId}" sent, got back id=${JSON.stringify(parsed.id)} (type: ${typeof parsed.id})`
1386
+ };
1387
+ });
1388
+ await test(
1389
+ "lifecycle-reinit-reject",
1390
+ "Rejects second initialize request",
1391
+ "lifecycle",
1392
+ false,
1393
+ "basic/lifecycle#initialization",
1394
+ async () => {
1395
+ try {
1396
+ const res = await rpc("initialize", {
1397
+ protocolVersion: SPEC_VERSION,
1398
+ capabilities: {},
1399
+ clientInfo: { name: "mcp-compliance", version: TOOL_VERSION }
1400
+ });
1401
+ const error = res.body?.error;
1402
+ if (error) {
1403
+ return { passed: true, details: `Re-initialization rejected with error: ${error.code} \u2014 ${error.message}` };
1404
+ }
1405
+ if (res.statusCode >= 400) {
1406
+ return { passed: true, details: `HTTP ${res.statusCode} (re-initialization rejected)` };
1407
+ }
1408
+ return {
1409
+ passed: false,
1410
+ details: `Server accepted second initialize (HTTP ${res.statusCode}) \u2014 should reject duplicate initialization`
1411
+ };
1412
+ } catch {
1413
+ return { passed: true, details: "Connection rejected (acceptable)" };
1414
+ }
1415
+ }
1416
+ );
1198
1417
  const hasLogging = !!serverInfo.capabilities.logging;
1199
1418
  await test(
1200
1419
  "lifecycle-logging",
@@ -1269,7 +1488,7 @@ async function runComplianceSuite(url, options = {}) {
1269
1488
  );
1270
1489
  await test(
1271
1490
  "lifecycle-progress",
1272
- "Accepts progress notifications",
1491
+ "Handles progress notifications gracefully",
1273
1492
  "lifecycle",
1274
1493
  false,
1275
1494
  "basic/utilities#progress",
@@ -1282,9 +1501,12 @@ async function runComplianceSuite(url, options = {}) {
1282
1501
  timeout
1283
1502
  );
1284
1503
  if (res.statusCode >= 200 && res.statusCode < 300) {
1285
- return { passed: true, details: `HTTP ${res.statusCode} (progress notification accepted)` };
1504
+ return { passed: true, details: `HTTP ${res.statusCode} (notification handled gracefully)` };
1286
1505
  }
1287
- return { passed: false, details: `HTTP ${res.statusCode} \u2014 server should accept progress notifications` };
1506
+ return {
1507
+ passed: false,
1508
+ details: `HTTP ${res.statusCode} \u2014 server should accept unknown notifications without error`
1509
+ };
1288
1510
  }
1289
1511
  );
1290
1512
  await test(
@@ -1326,7 +1548,11 @@ async function runComplianceSuite(url, options = {}) {
1326
1548
  return { passed: true, details: "HTTP 202 Accepted (correct)" };
1327
1549
  }
1328
1550
  if (res.statusCode >= 200 && res.statusCode < 300) {
1329
- return { passed: true, details: `HTTP ${res.statusCode} (accepted, but 202 is preferred)` };
1551
+ warnings.push(`Notification returned HTTP ${res.statusCode} instead of spec-required 202 Accepted`);
1552
+ return {
1553
+ passed: false,
1554
+ details: `HTTP ${res.statusCode} \u2014 spec requires 202 Accepted for notifications (MUST)`
1555
+ };
1330
1556
  }
1331
1557
  return { passed: false, details: `HTTP ${res.statusCode} \u2014 expected 202 Accepted for notifications` };
1332
1558
  }
@@ -1357,6 +1583,34 @@ async function runComplianceSuite(url, options = {}) {
1357
1583
  return { passed: false, details: `HTTP ${res.statusCode}` };
1358
1584
  }
1359
1585
  );
1586
+ await test(
1587
+ "transport-session-invalid",
1588
+ "Returns 404 for unknown session ID",
1589
+ "transport",
1590
+ false,
1591
+ "basic/transports#streamable-http",
1592
+ async () => {
1593
+ if (!sessionId) {
1594
+ return { passed: true, details: "Server did not issue session ID (test not applicable)" };
1595
+ }
1596
+ const fakeHeaders = {
1597
+ ...userHeaders,
1598
+ "mcp-session-id": "invalid-nonexistent-session-id"
1599
+ };
1600
+ if (negotiatedProtocolVersion) fakeHeaders["mcp-protocol-version"] = negotiatedProtocolVersion;
1601
+ const res = await mcpRequest(backendUrl, "ping", void 0, createIdCounter(99915), fakeHeaders, timeout);
1602
+ if (res.statusCode === 404) {
1603
+ return { passed: true, details: "HTTP 404 for unknown session ID (correct per spec)" };
1604
+ }
1605
+ if (res.statusCode === 400) {
1606
+ return {
1607
+ passed: false,
1608
+ details: "HTTP 400 \u2014 spec requires 404 (Not Found) for unrecognized session IDs, not 400"
1609
+ };
1610
+ }
1611
+ return { passed: false, details: `HTTP ${res.statusCode} \u2014 spec requires 404 for unrecognized MCP-Session-Id` };
1612
+ }
1613
+ );
1360
1614
  await test(
1361
1615
  "transport-get-stream",
1362
1616
  "GET with session returns SSE or 405",
@@ -1373,7 +1627,8 @@ async function runComplianceSuite(url, options = {}) {
1373
1627
  signal: AbortSignal.timeout(Math.min(timeout, 3e3))
1374
1628
  });
1375
1629
  const body = await res.body.text();
1376
- const ct = (res.headers["content-type"] || "").toLowerCase();
1630
+ const rawCt2 = res.headers["content-type"];
1631
+ const ct = (Array.isArray(rawCt2) ? rawCt2[0] : rawCt2 || "").toLowerCase();
1377
1632
  if (res.statusCode === 405) {
1378
1633
  return { passed: true, details: "HTTP 405 (server does not support server-initiated messages)" };
1379
1634
  }
@@ -1412,7 +1667,8 @@ async function runComplianceSuite(url, options = {}) {
1412
1667
  signal: AbortSignal.timeout(timeout)
1413
1668
  }).then(async (res) => {
1414
1669
  const text = await res.body.text();
1415
- const ct = (res.headers["content-type"] || "").toLowerCase();
1670
+ const rawCtConcurrent = res.headers["content-type"];
1671
+ const ct = (Array.isArray(rawCtConcurrent) ? rawCtConcurrent[0] : rawCtConcurrent || "").toLowerCase();
1416
1672
  let body;
1417
1673
  if (ct.includes("text/event-stream")) {
1418
1674
  body = parseSSEResponse(text);
@@ -1439,6 +1695,42 @@ async function runComplianceSuite(url, options = {}) {
1439
1695
  return { passed: true, details: `${results.length} concurrent requests handled correctly` };
1440
1696
  }
1441
1697
  );
1698
+ await test(
1699
+ "transport-sse-event-field",
1700
+ "SSE responses include event: message",
1701
+ "transport",
1702
+ false,
1703
+ "basic/transports#streamable-http",
1704
+ async () => {
1705
+ const res = await request(backendUrl, {
1706
+ method: "POST",
1707
+ headers: {
1708
+ "Content-Type": "application/json",
1709
+ Accept: "text/event-stream",
1710
+ ...buildHeaders()
1711
+ },
1712
+ body: JSON.stringify({ jsonrpc: "2.0", id: createIdCounter(99940)(), method: "ping" }),
1713
+ signal: AbortSignal.timeout(timeout)
1714
+ });
1715
+ const text = await res.body.text();
1716
+ const rawCtSse = res.headers["content-type"];
1717
+ const ct = (Array.isArray(rawCtSse) ? rawCtSse[0] : rawCtSse || "").toLowerCase();
1718
+ if (!ct.includes("text/event-stream")) {
1719
+ return { passed: true, details: "Server responded with JSON (not SSE) \u2014 event field check not applicable" };
1720
+ }
1721
+ const hasEventMessage = /^event:\s*message\s*$/m.test(text);
1722
+ if (hasEventMessage) {
1723
+ return { passed: true, details: "SSE response includes required event: message field" };
1724
+ }
1725
+ if (text.includes("data:")) {
1726
+ return {
1727
+ passed: false,
1728
+ details: "SSE response has data: fields but missing required event: message field (spec: MUST include event: message)"
1729
+ };
1730
+ }
1731
+ return { passed: true, details: "SSE response empty or no data fields \u2014 check not applicable" };
1732
+ }
1733
+ );
1442
1734
  const hasTools = !!serverInfo.capabilities.tools;
1443
1735
  let cachedToolsList = null;
1444
1736
  await test(
@@ -1520,9 +1812,6 @@ async function runComplianceSuite(url, options = {}) {
1520
1812
  issues.push(`${tool.name}: annotations.${field} should be boolean, got ${typeof ann[field]}`);
1521
1813
  }
1522
1814
  }
1523
- if (ann.title !== void 0 && typeof ann.title !== "string") {
1524
- issues.push(`${tool.name}: annotations.title should be string`);
1525
- }
1526
1815
  }
1527
1816
  if (issues.length > 0) return { passed: false, details: issues.join("; ") };
1528
1817
  return {
@@ -2070,6 +2359,63 @@ async function runComplianceSuite(url, options = {}) {
2070
2359
  }
2071
2360
  return { passed: false, details: `HTTP ${res.statusCode} \u2014 expected error code -32600` };
2072
2361
  });
2362
+ const undeclaredMethods = [
2363
+ { method: "tools/list", capability: "tools", declared: hasTools },
2364
+ { method: "resources/list", capability: "resources", declared: hasResources },
2365
+ { method: "prompts/list", capability: "prompts", declared: hasPrompts }
2366
+ ];
2367
+ const undeclared = undeclaredMethods.filter((m) => !m.declared);
2368
+ await test(
2369
+ "error-capability-gated",
2370
+ "Rejects methods for undeclared capabilities",
2371
+ "errors",
2372
+ false,
2373
+ "basic/lifecycle#capability-negotiation",
2374
+ async () => {
2375
+ if (undeclared.length === 0) {
2376
+ return {
2377
+ passed: true,
2378
+ details: "Server declares all capabilities (tools, resources, prompts) \u2014 no undeclared methods to test"
2379
+ };
2380
+ }
2381
+ const issues = [];
2382
+ for (const { method, capability } of undeclared) {
2383
+ const res = await rpc(method);
2384
+ const error = res.body?.error;
2385
+ if (!error && res.body?.result) {
2386
+ issues.push(`${method} returned success despite missing ${capability} capability`);
2387
+ }
2388
+ }
2389
+ if (issues.length > 0) return { passed: false, details: issues.join("; ") };
2390
+ return {
2391
+ passed: true,
2392
+ details: `Tested ${undeclared.length} undeclared method(s): ${undeclared.map((m) => m.method).join(", ")} \u2014 all returned errors`
2393
+ };
2394
+ }
2395
+ );
2396
+ const listMethodForCursor = hasTools ? "tools/list" : hasResources ? "resources/list" : hasPrompts ? "prompts/list" : null;
2397
+ await test(
2398
+ "error-invalid-cursor",
2399
+ "Handles invalid pagination cursor gracefully",
2400
+ "errors",
2401
+ false,
2402
+ "basic",
2403
+ async () => {
2404
+ if (!listMethodForCursor) {
2405
+ return { passed: true, details: "No list methods available to test (skipped)" };
2406
+ }
2407
+ const res = await rpc(listMethodForCursor, { cursor: "!!!invalid-garbage-cursor-$$$" });
2408
+ const error = res.body?.error;
2409
+ if (error) {
2410
+ return { passed: true, details: `Invalid cursor rejected with error: ${error.code} \u2014 ${error.message}` };
2411
+ }
2412
+ const result = res.body?.result;
2413
+ if (result) {
2414
+ return { passed: true, details: "Server returned results (likely ignored invalid cursor)" };
2415
+ }
2416
+ return { passed: false, details: "No error or result for invalid cursor" };
2417
+ }
2418
+ );
2073
2419
  const hasAuth = !!userHeaders.Authorization || !!userHeaders.authorization;
2074
2420
  await test(
2075
2421
  "security-auth-required",
@@ -2214,7 +2560,7 @@ async function runComplianceSuite(url, options = {}) {
2214
2560
  );
2215
2561
  await test(
2216
2562
  "security-oauth-metadata",
2217
- "OAuth metadata endpoint exists",
2563
+ "Protected Resource Metadata endpoint exists",
2218
2564
  "security",
2219
2565
  false,
2220
2566
  "basic/authorization",
@@ -2223,9 +2569,9 @@ async function runComplianceSuite(url, options = {}) {
2223
2569
  return { passed: true, details: "Skipped: server does not require auth" };
2224
2570
  }
2225
2571
  const parsedUrl = new URL(url);
2226
- const metadataUrl = `${parsedUrl.protocol}//${parsedUrl.host}/.well-known/oauth-authorization-server`;
2572
+ const prmUrl = `${parsedUrl.protocol}//${parsedUrl.host}/.well-known/oauth-protected-resource`;
2227
2573
  try {
2228
- const res = await request(metadataUrl, {
2574
+ const res = await request(prmUrl, {
2229
2575
  method: "GET",
2230
2576
  headers: { Accept: "application/json" },
2231
2577
  signal: AbortSignal.timeout(Math.min(timeout, 5e3))
@@ -2234,20 +2580,51 @@ async function runComplianceSuite(url, options = {}) {
2234
2580
  if (res.statusCode === 200) {
2235
2581
  try {
2236
2582
  const meta = JSON.parse(text);
2237
- if (meta.issuer && meta.token_endpoint) {
2238
- return { passed: true, details: `OAuth metadata found: issuer=${meta.issuer}` };
2583
+ if (!meta.resource) {
2584
+ return { passed: false, details: "PRM response missing required 'resource' field" };
2585
+ }
2586
+ if (!Array.isArray(meta.authorization_servers) || meta.authorization_servers.length === 0) {
2587
+ return { passed: false, details: "PRM response missing 'authorization_servers' array" };
2239
2588
  }
2240
2589
  return {
2241
- passed: false,
2242
- details: "OAuth metadata response missing required fields (issuer, token_endpoint)"
2590
+ passed: true,
2591
+ details: `Protected Resource Metadata found: resource=${meta.resource}, ${meta.authorization_servers.length} auth server(s)`
2243
2592
  };
2244
2593
  } catch {
2245
- return { passed: false, details: "OAuth metadata endpoint returned non-JSON response" };
2594
+ return { passed: false, details: "PRM endpoint returned non-JSON response" };
2246
2595
  }
2247
2596
  }
2248
- return { passed: false, details: `OAuth metadata endpoint returned HTTP ${res.statusCode}` };
2597
+ const legacyUrl = `${parsedUrl.protocol}//${parsedUrl.host}/.well-known/oauth-authorization-server`;
2598
+ try {
2599
+ const legacyRes = await request(legacyUrl, {
2600
+ method: "GET",
2601
+ headers: { Accept: "application/json" },
2602
+ signal: AbortSignal.timeout(Math.min(timeout, 5e3))
2603
+ });
2604
+ const legacyText = await legacyRes.body.text();
2605
+ if (legacyRes.statusCode === 200) {
2606
+ try {
2607
+ const legacyMeta = JSON.parse(legacyText);
2608
+ if (legacyMeta.issuer && legacyMeta.token_endpoint) {
2609
+ warnings.push(
2610
+ "Server uses legacy /.well-known/oauth-authorization-server instead of /.well-known/oauth-protected-resource (RFC 9728). Update to PRM for 2025-11-25 compliance."
2611
+ );
2612
+ return {
2613
+ passed: true,
2614
+ details: `Legacy OAuth AS metadata found: issuer=${legacyMeta.issuer} (should migrate to PRM)`
2615
+ };
2616
+ }
2617
+ } catch {
2618
+ }
2619
+ }
2620
+ } catch {
2621
+ }
2622
+ return {
2623
+ passed: false,
2624
+ details: `PRM endpoint returned HTTP ${res.statusCode} and no legacy OAuth metadata found`
2625
+ };
2249
2626
  } catch {
2250
- return { passed: false, details: "OAuth metadata endpoint unreachable" };
2627
+ return { passed: false, details: "PRM endpoint unreachable" };
2251
2628
  }
2252
2629
  }
2253
2630
  );
@@ -2323,6 +2700,44 @@ async function runComplianceSuite(url, options = {}) {
2323
2700
  }
2324
2701
  }
2325
2702
  );
2703
+ await test(
2704
+ "security-origin-validation",
2705
+ "Validates Origin header on requests",
2706
+ "security",
2707
+ false,
2708
+ "basic/transports#streamable-http",
2709
+ async () => {
2710
+ try {
2711
+ const res = await request(backendUrl, {
2712
+ method: "POST",
2713
+ headers: {
2714
+ "Content-Type": "application/json",
2715
+ Accept: "application/json, text/event-stream",
2716
+ Origin: "https://evil-rebinding-attack.example.com",
2717
+ ...buildHeaders()
2718
+ },
2719
+ body: JSON.stringify({ jsonrpc: "2.0", id: createIdCounter(99970)(), method: "ping" }),
2720
+ signal: AbortSignal.timeout(timeout)
2721
+ });
2722
+ await res.body.text();
2723
+ if (res.statusCode === 403 || res.statusCode === 401) {
2724
+ return { passed: true, details: `HTTP ${res.statusCode} (suspicious Origin rejected)` };
2725
+ }
2726
+ if (res.statusCode >= 200 && res.statusCode < 300) {
2727
+ return {
2728
+ passed: false,
2729
+ details: `HTTP ${res.statusCode} \u2014 server accepted request with untrusted Origin header (spec: MUST validate Origin for DNS rebinding protection)`
2730
+ };
2731
+ }
2732
+ if (res.statusCode >= 400) {
2733
+ return { passed: true, details: `HTTP ${res.statusCode} (suspicious Origin rejected)` };
2734
+ }
2735
+ return { passed: false, details: `HTTP ${res.statusCode}` };
2736
+ } catch {
2737
+ return { passed: true, details: "Connection rejected (acceptable)" };
2738
+ }
2739
+ }
2740
+ );
2326
2741
  async function runInjectionTest(toolName, paramName, payloads, detectPattern, label) {
2327
2742
  const issues = [];
2328
2743
  for (const payload of payloads) {
@@ -2843,13 +3258,13 @@ async function runComplianceSuite(url, options = {}) {
2843
3258
  function registerTools(server) {
2844
3259
  server.tool(
2845
3260
  "mcp_compliance_test",
2846
- "Run the full MCP compliance test suite against a server URL. Returns grade (A-F), score, and detailed results for all 69 tests covering transport, lifecycle, tools, resources, prompts, errors, schema validation, and security.",
3261
+ "Run the full MCP compliance test suite against a server URL. Returns grade (A-F), score, and detailed results for all 78 tests covering transport, lifecycle, tools, resources, prompts, errors, schema validation, and security.",
2847
3262
  {
2848
3263
  url: z.string().url().describe("The MCP server URL to test (must be HTTP or HTTPS)"),
2849
3264
  auth: z.string().optional().describe('Authorization header value (e.g., "Bearer tok123")'),
2850
3265
  headers: z.record(z.string()).optional().describe('Additional headers to include on all requests (e.g., {"X-Api-Key": "abc"})'),
2851
- timeout: z.number().optional().describe("Request timeout in milliseconds (default: 15000)"),
2852
- retries: z.number().optional().describe("Number of retries for failed tests (default: 0)"),
3266
+ timeout: z.number().int().min(1).max(3e5).optional().describe("Request timeout in milliseconds (default: 15000, max: 300000)"),
3267
+ retries: z.number().int().min(0).max(10).optional().describe("Number of retries for failed tests (default: 0, max: 10)"),
2853
3268
  only: z.array(z.string()).optional().describe("Only run tests matching these categories or test IDs"),
2854
3269
  skip: z.array(z.string()).optional().describe("Skip tests matching these categories or test IDs")
2855
3270
  },
@@ -2914,7 +3329,7 @@ ${JSON.stringify(report, null, 2)}` }
2914
3329
  url: z.string().url().describe("The MCP server URL to test"),
2915
3330
  auth: z.string().optional().describe('Authorization header value (e.g., "Bearer tok123")'),
2916
3331
  headers: z.record(z.string()).optional().describe("Additional headers to include on all requests"),
2917
- timeout: z.number().optional().describe("Request timeout in milliseconds (default: 15000)")
3332
+ timeout: z.number().int().min(1).max(3e5).optional().describe("Request timeout in milliseconds (default: 15000, max: 300000)")
2918
3333
  },
2919
3334
  {
2920
3335
  title: "Get Compliance Badge",
@@ -3265,7 +3680,9 @@ function parseList(value) {
3265
3680
  }
3266
3681
  var program = new Command();
3267
3682
  program.name("mcp-compliance").description("Test MCP servers for spec compliance").version(version2);
3268
- program.command("test").description("Run the full compliance test suite against an MCP server").argument("<url>", "MCP server URL to test").option("--format <format>", "Output format: terminal, json, or sarif", "terminal").option("--strict", "Exit with code 1 on any required test failure (for CI)").option("-H, --header <header>", 'Add header to all requests (format: "Key: Value", repeatable)', parseHeaderArg, {}).option("--auth <token>", 'Shorthand for -H "Authorization: <token>"').option("--timeout <ms>", "Request timeout in milliseconds", "15000").option("--retries <n>", "Number of retries for failed tests", "0").option(
3683
+ program.command("test").description("Run the full compliance test suite against an MCP server").argument("<url>", "MCP server URL to test").addOption(
3684
+ new Option("--format <format>", "Output format").choices(["terminal", "json", "sarif"]).default("terminal")
3685
+ ).option("--strict", "Exit with code 1 on any required test failure (for CI)").option("-H, --header <header>", 'Add header to all requests (format: "Key: Value", repeatable)', parseHeaderArg, {}).option("--auth <token>", 'Shorthand for -H "Authorization: <token>"').option("--timeout <ms>", "Request timeout in milliseconds", "15000").option("--retries <n>", "Number of retries for failed tests", "0").option(
3269
3686
  "--only <items>",
3270
3687
  'Only run matching categories or test IDs, comma-separated (e.g., "transport,lifecycle" or "transport-post,lifecycle-init")',
3271
3688
  parseList