fix(langchain): only create tools node for client tools

libs/langchain/src/agents/ReactAgent.ts

@@ -146,7 +146,7 @@ export class ReactAgent<
     /**
      * add single tool node for all tools
      */
-    if (toolClasses.length > 0) {
+    if (toolClasses.filter(isClientTool).length > 0) {
       const toolNode = new ToolNode(toolClasses.filter(isClientTool), {
         signal: this.options.signal,
       });
@@ -206,7 +206,7 @@ export class ReactAgent<
     /**
      * add edges for tools node
      */
-    if (toolClasses.length > 0) {
+    if (toolClasses.filter(isClientTool).length > 0) {
       if (shouldReturnDirect.size > 0) {
         allNodeWorkflows.addConditionalEdges(
           "tools",

libs/langchain/src/agents/middlewareAgent/ReactAgent.ts

@@ -227,7 +227,7 @@ export class ReactAgent<
     /**
      * add single tool node for all tools
      */
-    if (toolClasses.length > 0) {
+    if (toolClasses.filter(isClientTool).length > 0) {
       const toolNode = new ToolNode(toolClasses.filter(isClientTool), {
         signal: this.options.signal,
       });
@@ -354,7 +354,7 @@ export class ReactAgent<
     /**
      * add edges for tools node
      */
-    if (toolClasses.length > 0) {
+    if (toolClasses.filter(isClientTool).length > 0) {
       // Tools should return to first beforeModel node or agent
       let toolReturnTarget: string;
       if (beforeModelNodes.length > 0) {

libs/langchain/src/agents/middlewareAgent/tests/codeExecution.int.test.ts

@@ -0,0 +1,80 @@
+import { describe, it, expect } from "vitest";
+import { ChatAnthropic } from "@langchain/anthropic";
+import { HumanMessage, AIMessage } from "@langchain/core/messages";
+
+import { createAgent } from "../index.js";
+
+describe("Code Execution Tool", () => {
+  it("should use Anthropic code_execution tool to calculate statistics", async () => {
+    // Create ChatAnthropic model with code execution beta header
+    const model = new ChatAnthropic({
+      model: "claude-3-5-haiku-20241022",
+      temperature: 0,
+      clientOptions: {
+        defaultHeaders: {
+          "anthropic-beta": "code-execution-2025-08-25",
+        },
+      },
+    });
+
+    // Define the built-in code_execution tool
+    const codeExecutionTool = {
+      type: "code_execution_20250825",
+      name: "code_execution",
+    };
+
+    // Create agent with the built-in tool
+    const agent = createAgent({
+      model,
+      tools: [codeExecutionTool],
+    });
+
+    // Invoke the agent with a statistics calculation task
+    const result = await agent.invoke({
+      messages: [
+        new HumanMessage(
+          "Calculate the mean and standard deviation of [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]"
+        ),
+      ],
+    });
+
+    // Verify the result contains messages
+    expect(result.messages).toBeDefined();
+    expect(result.messages.length).toBeGreaterThan(0);
+
+    // Find the final AI response
+    const aiResponse = result.messages.find((msg) => msg.type === "ai") as
+      | AIMessage
+      | undefined;
+    expect(aiResponse).toBeDefined();
+
+    // The response should contain content blocks
+    const content = Array.isArray(aiResponse?.content)
+      ? aiResponse.content
+      : [];
+    expect(content.length).toBeGreaterThan(0);
+
+    // Verify that code_execution tool was used (server_tool_use block)
+    expect(content.some((block) => block.type === "server_tool_use")).toBe(
+      true
+    );
+
+    // Verify that we got a code execution result
+    expect(
+      content.some((block) => block.type === "bash_code_execution_tool_result")
+    ).toBe(true);
+
+    // The response should contain text with the calculated values
+    const textBlocks = content.filter((block) => block.type === "text");
+    const responseText = textBlocks
+      .map((block) => block.text)
+      .join(" ")
+      .toLowerCase();
+
+    // The response should mention the mean (5.5) and standard deviation (~2.87)
+    expect(responseText).toMatch(/mean|average/i);
+    expect(responseText).toMatch(/standard deviation|std/i);
+    expect(responseText).toMatch(/5\.5/); // Expected mean
+    expect(responseText).toMatch(/2\.[89]|3\./); // Expected stdev (approximately 2.87)
+  }, 60000); // Set timeout to 60s for API call
+});

libs/langchain/src/agents/tests/codeExecution.int.test.ts

@@ -0,0 +1,80 @@
+import { describe, it, expect } from "vitest";
+import { ChatAnthropic } from "@langchain/anthropic";
+import { HumanMessage, AIMessage } from "@langchain/core/messages";
+
+import { createAgent } from "../createAgent.js";
+
+describe("Code Execution Tool", () => {
+  it("should use Anthropic code_execution tool to calculate statistics", async () => {
+    // Create ChatAnthropic model with code execution beta header
+    const model = new ChatAnthropic({
+      model: "claude-3-5-haiku-20241022",
+      temperature: 0,
+      clientOptions: {
+        defaultHeaders: {
+          "anthropic-beta": "code-execution-2025-08-25",
+        },
+      },
+    });
+
+    // Define the built-in code_execution tool
+    const codeExecutionTool = {
+      type: "code_execution_20250825",
+      name: "code_execution",
+    };
+
+    // Create agent with the built-in tool
+    const agent = createAgent({
+      llm: model,
+      tools: [codeExecutionTool],
+    });
+
+    // Invoke the agent with a statistics calculation task
+    const result = await agent.invoke({
+      messages: [
+        new HumanMessage(
+          "Calculate the mean and standard deviation of [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]"
+        ),
+      ],
+    });
+
+    // Verify the result contains messages
+    expect(result.messages).toBeDefined();
+    expect(result.messages.length).toBeGreaterThan(0);
+
+    // Find the final AI response
+    const aiResponse = result.messages.find((msg) => msg.type === "ai") as
+      | AIMessage
+      | undefined;
+    expect(aiResponse).toBeDefined();
+
+    // The response should contain content blocks
+    const content = Array.isArray(aiResponse?.content)
+      ? aiResponse.content
+      : [];
+    expect(content.length).toBeGreaterThan(0);
+
+    // Verify that code_execution tool was used (server_tool_use block)
+    expect(content.some((block) => block.type === "server_tool_use")).toBe(
+      true
+    );
+
+    // Verify that we got a code execution result
+    expect(
+      content.some((block) => block.type === "bash_code_execution_tool_result")
+    ).toBe(true);
+
+    // The response should contain text with the calculated values
+    const textBlocks = content.filter((block) => block.type === "text");
+    const responseText = textBlocks
+      .map((block) => block.text)
+      .join(" ")
+      .toLowerCase();
+
+    // The response should mention the mean (5.5) and standard deviation (~2.87)
+    expect(responseText).toMatch(/mean|average/i);
+    expect(responseText).toMatch(/standard deviation|std/i);
+    expect(responseText).toMatch(/5\.5/); // Expected mean
+    expect(responseText).toMatch(/2\.[89]|3\./); // Expected stdev (approximately 2.87)
+  }, 60000); // Set timeout to 60s for API call
+});