Fixes #331: [Fleet Execution] Implement ToolFiltering Custom AI Workflow Example

packages/sdk/README.md

@@ -30,6 +30,10 @@ To use `stitchTools()` with the [Vercel AI SDK](https://sdk.vercel.ai/), install
 npm install @google/stitch-sdk ai
 ```
 
+## Examples
+
+See the `examples/` directory for sample code, including [Tool Filtering Custom AI Workflow Example](examples/tool-filtering/README.md).
+
 ## Working with Projects and Screens
 
 ### List existing projects

packages/sdk/examples/tool-filtering/README.md

@@ -0,0 +1,19 @@
+# Tool Filtering Custom AI Workflow Example
+
+This example demonstrates how to use the `stitchTools({ include: [...] })` functionality with the Vercel AI SDK to filter the available Stitch MCP tools.
+
+Filtering tools is useful for:
+- **Reducing context window usage** by only exposing the necessary tool schemas to the LLM.
+- **Enforcing least privilege** so an agent can only perform specific operations (e.g., read-only access to screens without the ability to create new projects or edit design systems).
+
+## Prerequisites
+
+- `STITCH_API_KEY` (or `STITCH_ACCESS_TOKEN`) for the Stitch SDK.
+- `GEMINI_API_KEY` (or another AI SDK provider token) for the LLM.
+
+## How to run
+
+```bash
+# Run the example script
+bun index.ts
+```

packages/sdk/examples/tool-filtering/index.ts

@@ -0,0 +1,65 @@
+// Copyright 2026 Google LLC
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     https://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+import { stitchTools } from "@google/stitch-sdk/ai";
+import { generateText } from "ai";
+import { createGoogleGenerativeAI } from "@ai-sdk/google";
+
+async function main() {
+  // Validate environment variables
+  if (!process.env.GEMINI_API_KEY) {
+    console.error("Error: GEMINI_API_KEY environment variable is required.");
+    process.exit(1);
+  }
+  if (!process.env.STITCH_API_KEY && !process.env.STITCH_ACCESS_TOKEN) {
+    console.error("Error: STITCH_API_KEY or STITCH_ACCESS_TOKEN environment variable is required.");
+    process.exit(1);
+  }
+
+  console.log("Configuring agent with restricted toolset...");
+
+  // Only provide the specific tools needed for this workflow to reduce
+  // context window usage and prevent the agent from using other tools.
+  const allowedTools = [
+    "generate_screen_from_text",
+    "get_screen",
+  ];
+
+  console.log(`Allowed tools: ${allowedTools.join(", ")}`);
+
+  const tools = stitchTools({ include: allowedTools });
+
+  const google = createGoogleGenerativeAI();
+  const model = google("gemini-2.5-pro");
+
+  console.log("\nPrompting agent to generate a screen...");
+  const result = await generateText({
+    model,
+    tools,
+    prompt: "Generate a screen for a modern analytics dashboard. Please provide the ID of the screen you generated.",
+    maxSteps: 5,
+  });
+
+  console.log("\nAgent Response:");
+  console.log(result.text);
+
+  console.log("\nTool calls made by agent:");
+  for (const step of result.steps) {
+    for (const toolCall of step.toolCalls) {
+       console.log(`- ${toolCall.toolName}`);
+    }
+  }
+}
+
+main().catch(console.error);