fix: forward workflow settings outside the typed create schema (v2.74.1) (czlonkowski#1027)

* fix: forward workflow settings outside the typed create schema (v2.74.1)

n8n_create_workflow declared settings as a closed z.object, so keys such as
availableInMCP, callerPolicy and timeSavedPerExecution were stripped before
the payload reached the API client, while updateSettings forwarded them.
The typed keys stay validated; everything else now passes through, as on the
update path. Tool schema and docs mention availableInMCP.

Fixes czlonkowski#1026

Conceived by Romuald Członkowski - www.aiadvisors.pl/en

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J5E2LQVUpMQDBDueBzieGp

* test: give the second memory node its own id in the multiple-memory fixture

n8n 2.36 rejects a workflow whose nodes share an id, so the live
integration test "should validate memory connections" has failed on main
since the 2.36 update. The helper defaults every memory node to
'memory-1'; the second one now gets 'memory-2'.

Conceived by Romuald Członkowski - www.aiadvisors.pl/en

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J5E2LQVUpMQDBDueBzieGp

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

Author: czlonkowski

CHANGELOG.md

@@ -7,6 +7,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [2.74.1] - 2026-08-27
+
+### Fixed
+
+- **`n8n_create_workflow` silently dropped workflow settings outside its own list** (#1026). The create input schema was a closed object with eight settings keys, so `availableInMCP`, `callerPolicy`, `callerIds`, `timeSavedPerExecution`, `customTelemetryTags`, `redactionPolicy` and `timeSavedMode` were stripped before the payload reached the API client. A workflow created with `settings.availableInMCP: true` therefore never appeared in n8n's instance-level MCP server, while the same key set through `updateSettings` worked. The typed keys are still validated; every other key is now forwarded, as on the update path; the create cleaner still drops the derived properties n8n ignores on write. The tool schema and documentation now mention `availableInMCP`.
+
 ## [2.74.0] - 2026-08-27
 
 ### Changed

package.json

@@ -1,6 +1,6 @@
 {
   "name": "n8n-mcp",
-  "version": "2.74.0",
+  "version": "2.74.1",
   "description": "Integration between n8n workflow automation and Model Context Protocol (MCP)",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package.runtime.json

@@ -1,6 +1,6 @@
 {
   "name": "n8n-mcp-runtime",
-  "version": "2.74.0",
+  "version": "2.74.1",
   "description": "n8n MCP Server Runtime Dependencies Only",
   "private": true,
   "dependencies": {

src/mcp/handlers-n8n-manager.ts

@@ -443,6 +443,9 @@ const createWorkflowSchema = z.object({
   // Two-arg z.record(keySchema, valueSchema) — see services/n8n-validation.ts for the
   // Zod 3/4 compatibility rationale (#744).
   connections: z.preprocess(normalizeMcpWorkflowConnections, z.record(z.string(), z.any())),
+  // The typed keys are validated; every other key is forwarded, as on the update path.
+  // A closed object here silently dropped `availableInMCP`, `callerPolicy` and the other
+  // settings added since n8n 1.119 before they reached the cleaner (issue #1026).
   settings: z.preprocess(normalizeMcpJsonValue, z.object({
     executionOrder: z.enum(['v0', 'v1']).optional(),
     timezone: z.string().optional(),
@@ -452,7 +455,7 @@ const createWorkflowSchema = z.object({
     saveExecutionProgress: z.boolean().optional(),
     executionTimeout: z.number().optional(),
     errorWorkflow: z.string().optional(),
-  })).optional(),
+  }).passthrough()).optional(),
   // Validated by parseNodeGroupsInput() — see services/node-groups.ts
   nodeGroups: z.any().optional(),
   projectId: z.string().optional(),

src/mcp/tool-docs/workflow_management/n8n-create-workflow.ts

@@ -21,7 +21,7 @@ export const n8nCreateWorkflowDoc: ToolDocumentation = {
       name: { type: 'string', required: true, description: 'Workflow name' },
       nodes: { type: 'array', required: true, description: 'Array of nodes with id, name, type, typeVersion, position, parameters' },
       connections: { type: 'object', required: true, description: 'Node connections. Keys are source node names (not IDs)' },
-      settings: { type: 'object', description: 'Optional workflow settings (timezone, error handling, etc.)' },
+      settings: { type: 'object', description: 'Optional workflow settings (timezone, error handling, etc.). Every key the n8n Public API accepts is forwarded, including availableInMCP, which exposes the workflow to n8n\'s instance-level MCP server.' },
       nodeGroups: { type: 'array', description: 'Optional canvas groups (n8n 2.28+): [{name, nodeIds, description?}]. Members are node IDs from nodes[] and must form a connected run with no trigger among them. Dropped with a warning on n8n older than 2.28.' },
       projectId: { type: 'string', description: 'Optional project to create the workflow in (enterprise feature). Defaults to the personal project.' },
       parentFolderId: { type: 'string', description: 'Optional folder to place the workflow in (n8n 2.32+; rejected with a 400 on older instances). Omit for the project root. Manage folders with n8n_manage_folders.' }
@@ -72,7 +72,8 @@ n8n_create_workflow({
     timezone: "America/New_York",
     errorWorkflow: "error_handler_workflow_id",
     saveDataSuccessExecution: "all",
-    saveDataErrorExecution: "all"
+    saveDataErrorExecution: "all",
+    availableInMCP: true  // expose to n8n's instance-level MCP server
   }
 })`
     ],

src/mcp/tool-docs/workflow_management/n8n-update-partial-workflow.ts

@@ -50,7 +50,7 @@ export const n8nUpdatePartialWorkflowDoc: ToolDocumentation = {
 - **replaceConnections**: Replace entire connections object
 
 ### Metadata Operations (5 types):
-- **updateSettings**: Modify workflow settings
+- **updateSettings**: Modify workflow settings (merged over the current ones). Any key the n8n Public API accepts is forwarded, e.g. \`availableInMCP: true\` exposes the workflow to n8n's instance-level MCP server.
 - **updateName**: Rename the workflow
 - **setNodeGroups**: Replace the workflow's canvas groups (n8n 2.28+). Full replacement — pass every group to keep, or \`[]\` to ungroup everything. Each group takes \`name\` plus either \`nodeNames\` or \`nodeIds\`, and an optional \`description\` (max 155 chars, n8n 2.32+; dropped automatically on older instances). Group members must form a connected run with no trigger among them; n8n validates that on save and its message is returned unchanged if a group you asked for is rejected.
 - **addTag**: Add a workflow tag

src/mcp/tools-n8n-manager.ts

@@ -52,7 +52,7 @@ export const n8nManagementTools: ToolDefinition[] = [
         },
         settings: {
           type: 'object',
-          description: 'Optional workflow settings (execution order, timezone, error handling)',
+          description: 'Optional workflow settings (execution order, timezone, error handling). Any other key the n8n Public API accepts is forwarded as well, e.g. availableInMCP (expose the workflow to n8n\'s instance-level MCP server), callerPolicy, callerIds.',
           properties: {
             executionOrder: { type: 'string', enum: ['v0', 'v1'] },
             timezone: { type: 'string' },
@@ -61,7 +61,8 @@ export const n8nManagementTools: ToolDefinition[] = [
             saveManualExecutions: { type: 'boolean' },
             saveExecutionProgress: { type: 'boolean' },
             executionTimeout: { type: 'number' },
-            errorWorkflow: { type: 'string' }
+            errorWorkflow: { type: 'string' },
+            availableInMCP: { type: 'boolean', description: 'Expose the workflow to n8n\'s instance-level MCP server (n8n 1.119+)' }
           }
         },
         nodeGroups: {

tests/integration/ai-validation/ai-agent-validation.test.ts

@@ -317,6 +317,7 @@ describe('Integration: AI Agent Validation', () => {
     });
 
     const memory2 = createMemoryNode({
+      id: 'memory-2', // the helper defaults to 'memory-1'; n8n 2.36 rejects duplicate node ids
       name: 'Memory 2'
     });
 

tests/unit/mcp/handlers-n8n-manager.test.ts

@@ -308,6 +308,45 @@ describe('handlers-n8n-manager', () => {
       expect(n8nValidation.validateWorkflowStructure).toHaveBeenCalledWith(input);
     });
 
+    it('should forward settings keys outside the typed schema, such as availableInMCP (issue #1026)', async () => {
+      // Regression: the create schema was a closed z.object, so Zod stripped every settings key
+      // it did not list before the payload reached the API client. The update path forwards them.
+      const testWorkflow = createTestWorkflow();
+      const input = {
+        name: 'Test Workflow',
+        nodes: testWorkflow.nodes,
+        connections: testWorkflow.connections,
+        settings: {
+          executionOrder: 'v1',
+          availableInMCP: true,
+          callerPolicy: 'workflowsFromSameOwner',
+          timeSavedPerExecution: 5,
+        },
+      };
+
+      mockApiClient.createWorkflow.mockResolvedValue(testWorkflow);
+
+      const result = await handlers.handleCreateWorkflow(input);
+
+      expect(result.success).toBe(true);
+      const sentWorkflow = mockApiClient.createWorkflow.mock.calls[0][0];
+      expect(sentWorkflow.settings).toEqual(input.settings);
+    });
+
+    it('should still reject invalid values for the typed settings keys', async () => {
+      const testWorkflow = createTestWorkflow();
+      const result = await handlers.handleCreateWorkflow({
+        name: 'Test Workflow',
+        nodes: testWorkflow.nodes,
+        connections: testWorkflow.connections,
+        settings: { executionOrder: 'v2', availableInMCP: true },
+      });
+
+      expect(result.success).toBe(false);
+      expect(result.error).toBe('Invalid input');
+      expect(mockApiClient.createWorkflow).not.toHaveBeenCalled();
+    });
+
     it('forwards parentFolderId to the create payload (folder placement, n8n 2.32+)', async () => {
       const testWorkflow = createTestWorkflow();
       const input = {