@intangle/mcp-server 2.3.7 → 2.3.9

package/dist/index.js

@@ -137,9 +137,8 @@ try {
         if (!space_id) {
             throw new Error("space_id is required. Use list_spaces to see available options.");
         }
-        return makeApiCall("search-memories", {
+        return makeApiCall("search", {
             space_id,
-            space_ids: [space_id], // Convert to array for backend compatibility
             query,
             topics,
         });

package/dist/tool-definitions.js

@@ -164,7 +164,7 @@ export const TOOLS = [
                 },
                 add: {
                     type: "object",
-                    description: "Add new items to context with automatic intelligent classification and topic suggestion",
+                    description: "Add new items. ALWAYS pass 'type' field on each item to skip expensive auto-classification.",
                     properties: {
                         items: {
                             type: "array",
@@ -174,7 +174,12 @@ export const TOOLS = [
                                     title: { type: "string", description: "Item title" },
                                     content: {
                                         type: "string",
-                                        description: "Item content - system automatically classifies as task (actionable), context (knowledge), or skill (workflow/procedure) and suggests relevant topics. VALUABLE CONTENT INCLUDES: decisions and their reasoning, user preferences observed, lessons learned, process descriptions, meeting outcomes, project context, expertise areas, working patterns, and anything that would help future sessions understand this user/space better."
+                                        description: "Item content. VALUABLE CONTENT INCLUDES: decisions and their reasoning, user preferences observed, lessons learned, process descriptions, meeting outcomes, project context, expertise areas, working patterns, and anything that would help future sessions understand this user/space better."
+                                    },
+                                    type: {
+                                        type: "string",
+                                        enum: ["task", "context", "skill", "document"],
+                                        description: "REQUIRED: Item type. 'task' for actionable items, 'context' for knowledge/facts, 'skill' for workflows, 'document' for long-form reference material. Omitting this triggers expensive auto-classification."
                                     },
                                     subtasks: {
                                         type: "array",
@@ -212,7 +217,7 @@ export const TOOLS = [
                                 },
                                 required: ["title", "content"]
                             },
-                            description: "Array of items to add. System intelligently: (1) classifies as task, context, or skill, (2) suggests 1-3 relevant topics based on content and existing topics. Examples: 'Need to fix auth bug' → task with topics like 'authentication', 'bug-fix'. 'Learned React batches updates' → context with topics like 'react', 'learning'. Tasks can include subtasks array for hierarchical task management."
+                            description: "Array of items to add. ALWAYS pass 'type' field to skip auto-classification. System suggests 1-3 relevant topics based on content. Tasks can include subtasks array for hierarchical task management."
                         }
                     }
                 },

package/index.ts

@@ -186,9 +186,8 @@ try {
       )
     }
 
-    return makeApiCall("search-memories", {
+    return makeApiCall("search", {
       space_id,
-      space_ids: [space_id], // Convert to array for backend compatibility
       query,
       topics,
     })

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@intangle/mcp-server",
-	"version": "2.3.7",
+	"version": "2.3.9",
 	"description": "Model Context Protocol server for Intangle - AI context that persists across conversations",
 	"main": "dist/index.js",
 	"type": "module",

package/tool-definitions.ts

@@ -187,7 +187,7 @@ export const TOOLS = [
         add: {
           type: "object",
           description:
-            "Add new items to context with automatic intelligent classification and topic suggestion",
+            "Add new items. ALWAYS pass 'type' field on each item to skip expensive auto-classification.",
           properties: {
             items: {
               type: "array",
@@ -198,7 +198,13 @@ export const TOOLS = [
                   content: {
                     type: "string",
                     description:
-                      "Item content - system automatically classifies as task (actionable), context (knowledge), or skill (workflow/procedure) and suggests relevant topics. VALUABLE CONTENT INCLUDES: decisions and their reasoning, user preferences observed, lessons learned, process descriptions, meeting outcomes, project context, expertise areas, working patterns, and anything that would help future sessions understand this user/space better."
+                      "Item content. VALUABLE CONTENT INCLUDES: decisions and their reasoning, user preferences observed, lessons learned, process descriptions, meeting outcomes, project context, expertise areas, working patterns, and anything that would help future sessions understand this user/space better."
+                  },
+                  type: {
+                    type: "string",
+                    enum: ["task", "context", "skill", "document"],
+                    description:
+                      "REQUIRED: Item type. 'task' for actionable items, 'context' for knowledge/facts, 'skill' for workflows, 'document' for long-form reference material. Omitting this triggers expensive auto-classification."
                   },
                   subtasks: {
                     type: "array",
@@ -242,7 +248,7 @@ export const TOOLS = [
                 required: ["title", "content"]
               },
               description:
-                "Array of items to add. System intelligently: (1) classifies as task, context, or skill, (2) suggests 1-3 relevant topics based on content and existing topics. Examples: 'Need to fix auth bug' → task with topics like 'authentication', 'bug-fix'. 'Learned React batches updates' → context with topics like 'react', 'learning'. Tasks can include subtasks array for hierarchical task management."
+                "Array of items to add. ALWAYS pass 'type' field to skip auto-classification. System suggests 1-3 relevant topics based on content. Tasks can include subtasks array for hierarchical task management."
             }
           }
         },