### Install and Run Project

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Install dependencies and start the development server. This is a common setup step for the project.

```bash
pnpm install
pnpm dev
```

--------------------------------

### Install Dependencies and Setup Environment

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-b2b-okr-manager/README.md

Clone the repository, install npm dependencies, and copy environment template files. This sets up the project locally for development.

```bash
git clone https://github.com/cloudflare/agents.git
cd agents/demos/mcp-stytch-b2b-okr-manager
npm i
```

```bash
cp .env.template .env.local
cp .dev.vars.template .dev.vars
```

--------------------------------

### Initialize and Start the MCP Server

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-client/README.md

Install dependencies and launch the MCP server environment.

```sh
npm install
npm start
```

--------------------------------

### Initialize and Run Local Development

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server/README.md

Commands to clone the repository, install dependencies using pnpm, and start the development server.

```bash
# clone the repository
git clone https://github.com/cloudflare/ai.git
# Or if using ssh:
# git clone git@github.com:cloudflare/ai.git

# install dependencies
cd ai
# Note: using pnpm instead of just "npm"
pnpm install

# run locally
npx nx dev remote-mcp-server
```

--------------------------------

### Initialize and run the MCP server locally

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server/static/README.md

Commands to clone the repository, install dependencies, and start the development server.

```bash
# clone the repository
git clone git@github.com:cloudflare/ai.git

# install dependencies
cd ai
npm install

# run locally
npx nx dev remote-mcp-server
```

--------------------------------

### Start Project Locally

Source: https://github.com/cloudflare/ai/blob/main/demos/orchestrator-workers/README.md

Use this command to start the project in development mode locally. Ensure you have Node.js and npm installed.

```bash
npx nx dev orchestrator-workers
```

--------------------------------

### Clone and run the project locally

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-descope-auth/static/README.md

Commands to clone the repository, install dependencies, and start the development server.

```bash
# clone the repository
git clone git@github.com:cloudflare/ai.git

# install dependencies
cd ai
npm install

# run locally
npx nx dev remote-mcp-server-descope-auth
```

--------------------------------

### Start Application

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager/README.md

Starts the application in development mode.

```bash
npx nx start agent-task-manager
```

--------------------------------

### Install Dependencies for Development

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-slack-oauth/README.md

Install project dependencies using npm. This command should be run before starting the development server or building the project.

```bash
npm install
```

--------------------------------

### Start Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager-human-in-the-loop/README.md

Use this command to start the project locally. It utilizes Nx for development.

```bash
npx nx dev agent-task-manager-human-in-the-loop
```

--------------------------------

### Initialize and Run Cloudflare AI Projects

Source: https://github.com/cloudflare/ai/blob/main/README.md

Commands for cloning the repository, installing dependencies, running examples, and executing tests.

```bash
# Clone and install
git clone git@github.com:cloudflare/ai.git
cd ai
pnpm install

# Run an example
cd examples/workers-ai
pnpm dev

# Run tests for a package
cd packages/workers-ai-provider
pnpm test

# Run E2E tests (requires Cloudflare credentials)
pnpm test:e2e
```

--------------------------------

### Start Local Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-scheduler/README.md

Use this command to start the project locally for development.

```bash
npx nx dev agent-scheduler
```

--------------------------------

### Start Development Server (NPM Script)

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-scheduler/README.md

Starts the development server using an NPM script.

```bash
npx nx start agent-scheduler
```

--------------------------------

### Clone and Install Dependencies

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-consumer-todo-list/README.md

Initializes the project directory and installs required Node.js packages.

```bash
git clone https://github.com/cloudflare/ai.git
cd ai
npm i
cd demos/mcp-stytch-consumer-todo-list
```

--------------------------------

### Clone and Install Dependencies for Remote MCP Server

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-server-bearer-auth/static/README.md

Clone the repository and install project dependencies using npm. This is the initial setup step for local development.

```bash
git clone git@github.com:cloudflare/ai.git

cd ai
npm install
```

--------------------------------

### Start Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/routing/README.md

Initiates the local development environment using Nx and Wrangler.

```bash
npx nx dev routing
```

--------------------------------

### Copy Environment Variables Template

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Copy the example environment variables file to be filled with your credentials. This is part of the setup for deploying the project.

```bash
cp .dev.vars.example .dev.vars
```

--------------------------------

### Start Local Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/prompt-chaining/README.md

Use this command to start the project locally for development. It utilizes Nx and the development server.

```bash
npx nx dev prompt-chaining
```

--------------------------------

### Start Project Locally

Source: https://github.com/cloudflare/ai/blob/main/demos/model-scraper/README.md

Command to initiate the project in a local development environment.

```bash
npx nx start model-scraper
```

--------------------------------

### Start Application in Development Mode

Source: https://github.com/cloudflare/ai/blob/main/demos/orchestrator-workers/README.md

Starts the application in development mode. This is similar to the 'dev' command and is useful for local testing.

```bash
npx nx start orchestrator-workers
```

--------------------------------

### Start Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling-stream-traditional/README.md

Use this command to start the development server locally. It utilizes Wrangler for managing Cloudflare Workers.

```bash
npx nx dev tool-calling-stream-traditional
```

--------------------------------

### Start development server

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling-stream/README.md

Run the project locally in development mode using Wrangler.

```bash
npx nx dev tool-calling-stream
```

--------------------------------

### Install Dependencies

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-cf-access-self-hosted/README.md

Install project dependencies using npm.

```bash
npm install
```

--------------------------------

### Start Development Server (Alias)

Source: https://github.com/cloudflare/ai/blob/main/demos/prompt-chaining/README.md

This NPM script is an alias for the 'dev' command, starting the development server.

```bash
npx nx start prompt-chaining
```

--------------------------------

### Install Dependencies

Source: https://github.com/cloudflare/ai/blob/main/packages/workers-ai-provider/README.md

Install the necessary packages for the Workers AI provider and the AI SDK.

```bash
npm install workers-ai-provider ai
```

--------------------------------

### Clone Repository and Install Dependencies

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-descope-auth/README.md

Clone the project repository, navigate into the directory, and install all necessary npm dependencies to set up the local development environment.

```bash
# clone the repository
git clone git@github.com:cloudflare/ai.git

# install dependencies
cd ai
npm install
```

--------------------------------

### Install AI Gateway Provider

Source: https://github.com/cloudflare/ai/blob/main/packages/ai-gateway-provider/README.md

Install the package via npm.

```bash
npm install ai-gateway-provider
```

--------------------------------

### Start Development Server with Wrangler

Source: https://github.com/cloudflare/ai/blob/main/demos/evaluator-optimiser/README.md

Starts the development server locally using Wrangler. This is useful for local testing and development.

```bash
npx nx dev evaluator-optimiser
```

--------------------------------

### AI Search Setup

Source: https://github.com/cloudflare/ai/blob/main/packages/workers-ai-provider/README.md

Connect your data to Cloudflare's managed RAG service for natural language querying. This example shows how to configure and use AI Search.

```APIDOC
## AI Search Setup

### Description
Configure and use Cloudflare's managed RAG service, AI Search, to connect and query your data using natural language.

### Configuration (`wrangler.jsonc`)
```jsonc
{
	"ai_search": [{ "binding": "AI_SEARCH", "name": "my-search-index" }]
}
```

### Usage (`.ts`)
```ts
import { createAISearch } from "workers-ai-provider";
import { generateText } from "ai";

const aisearch = createAISearch({ binding: env.AI_SEARCH });

const { text } = await generateText({
	model: aisearch(),
	messages: [{ role: "user", content: "How do I setup AI Gateway?" }],
});
```

**Note:** `createAutoRAG` is deprecated. Use `createAISearch` instead.
```

--------------------------------

### Start Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/ui-worker/README.md

Command to initiate the local development environment for the UI Worker project.

```bash
npx nx dev ui-worker
```

--------------------------------

### Install @cloudflare/tanstack-ai

Source: https://github.com/cloudflare/ai/blob/main/packages/tanstack-ai/README.md

Install the core @cloudflare/tanstack-ai and @tanstack/ai packages using npm.

```bash
npm install @cloudflare/tanstack-ai @tanstack/ai
```

--------------------------------

### Install Third-Party Model Plugins

Source: https://github.com/cloudflare/ai/blob/main/packages/workers-ai-provider/README.md

Install optional peer dependency plugins for specific third-party model providers to enable routing through AI Gateway.

```bash
npm install @ai-sdk/openai      # openai, deepseek, xai/grok, groq, mistral, perplexity, cerebras, openrouter, fireworks, alibaba, minimax
npm install @ai-sdk/anthropic   # anthropic
npm install @ai-sdk/google      # google, google-vertex
```

--------------------------------

### Start Agent Locally

Source: https://github.com/cloudflare/ai/blob/main/demos/hello-world/README.md

Initiates the local development server for the Agent.

```shell
npm start
```

--------------------------------

### Run Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager/README.md

Starts the local development server for the agent task manager.

```bash
npx nx dev agent-task-manager
```

--------------------------------

### Install AI Gateway Provider SDKs

Source: https://github.com/cloudflare/ai/blob/main/packages/tanstack-ai/README.md

Install specific SDKs for third-party AI providers when using AI Gateway.

```bash
# For OpenAI
npm install @tanstack/ai-openai

# For Anthropic
npm install @tanstack/ai-anthropic

# For Gemini
npm install @tanstack/ai-gemini

# For Grok
npm install @tanstack/ai-grok

# For OpenRouter
npm install @tanstack/ai-openrouter @openrouter/sdk
```

--------------------------------

### Start Development Server (Alias)

Source: https://github.com/cloudflare/ai/blob/main/demos/evaluator-optimiser/README.md

An alias for the 'dev' command, this starts the development server with Wrangler. It's a convenient shortcut for local development.

```bash
npx nx start evaluator-optimiser
```

--------------------------------

### Run Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-consumer-todo-list/README.md

Starts the local development environment and the MCP inspector tool.

```bash
npm run dev
```

```bash
npx @modelcontextprotocol/inspector@latest
```

--------------------------------

### POST /query API Request Example

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-scheduler/README.md

Example cURL command to send a query to the /query endpoint for processing scheduling actions.

```bash
curl -X POST \
  -H "Content-Type: application/json" \
  -d '{"agentId": "your-agent-id", "prompt": "your-query"}' \
  http://localhost:8787/query
```

--------------------------------

### Configure Environment Variables

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-consumer-todo-list/README.md

Example content for local environment files required for Stytch integration.

```text
# This is what a completed .env.local file will look like
VITE_STYTCH_PUBLIC_TOKEN=public-token-test-abc123-abcde-1234-0987-0000-abcd1234
VITE_STYTCH_DOMAIN=random-domain-name.customers.stytch.dev
```

```text
// This is what a completed .dev.vars file will look like
STYTCH_PROJECT_ID=project-test-6c20cd16-73d5-44f7-852c-9a7e7b2ccf62
```

--------------------------------

### Example cURL Command for API Interaction

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling-stream-traditional/README.md

This cURL command demonstrates how to send a POST request to the API endpoint to get weather information for a specified location.

```bash
curl -X POST \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What's the weather in London?"}' \
  http://localhost:8787/
```

--------------------------------

### Start Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-slack-oauth/README.md

Run the development server for local testing and debugging. This command typically uses a tool like Wrangler to serve the worker locally.

```bash
npm run dev
```

--------------------------------

### Run Local Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-autorag/static/README.md

Start the remote MCP server locally for development and testing.

```bash
npx nx dev remote-mcp-server-autorag
```

--------------------------------

### Configure Redirect URI

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-logto/README.md

Example format for the production callback URL to be added to Logto.

```text
https://remote-mcp-logto.<your-subdomain>.workers.dev/callback
```

--------------------------------

### Run Integration Tests for Tool Calling

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling/README.md

Execute the integration tests for the tool-calling project. Ensure all dependencies are installed before running.

```bash
npx nx test tool-calling
```

--------------------------------

### API Request Body Example

Source: https://github.com/cloudflare/ai/blob/main/demos/structured-output/README.md

Example JSON payload for the POST request to the local development server. This body includes a prompt for content generation.

```json
{
	"prompt": "Create a recipe for sourdough bread."
}
```

--------------------------------

### POST /confirmations/:confirmationId API Request Example

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-scheduler/README.md

Example cURL command to confirm or reject a pending scheduling operation via the /confirmations/:confirmationId endpoint.

```bash
curl -X POST \
  -H "Content-Type: application/json" \
  -d '{"agentId": "your-agent-id", "confirm": true}' \
  http://localhost:8787/confirmations/your-confirmation-id
```

--------------------------------

### Local Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-cf-access-self-hosted/README.md

Start the Cloudflare Worker locally for development using Wrangler.

```bash
wrangler dev
```

--------------------------------

### Run Local MCP Server

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-autorag/README.md

Start the remote MCP server locally for development and testing.

```bash
npx nx dev remote-mcp-server
```

--------------------------------

### Develop Python Worker with uv

Source: https://github.com/cloudflare/ai/blob/main/demos/python-workers-mcp/README.md

Use this command to start the development server for your Python Worker.

```console
uv run pywrangler dev
```

--------------------------------

### Run Local Development Server

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-descope-auth/README.md

Start the local development server using the Nx CLI. This command will build and run the remote MCP server on a specified port.

```bash
npx nx dev remote-mcp-server-descope-auth
```

--------------------------------

### Interact with API via Curl

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager/README.md

Example command to send a prompt to the task manager agent.

```bash
curl -X POST \
-H "Content-Type: application/json" \
-d '{"agentId": "your-agent-id", "prompt": "your-prompt"}' \
http://localhost:8787/
```

--------------------------------

### Test API with curl

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-auth0/todos-api/README.md

Verify the API by sending a GET request with an Authorization header.

```bash
curl --request GET \
  --url http://localhost:8788/api/me \
  --header 'Authorization: Bearer eyJhbGciOiJSUzI1NiIsInR5cCI6Im...'
```

--------------------------------

### Deploy Application with Wrangler

Source: https://github.com/cloudflare/ai/blob/main/demos/evaluator-optimiser/README.md

Use this command to deploy the application using Wrangler. Ensure you have Wrangler installed and configured.

```bash
npx nx deploy evaluator-optimiser
```

--------------------------------

### Configure Cursor MCP command

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-github-oauth/README.md

Example command format for connecting Cursor to a remote MCP server.

```bash
npx mcp-remote https://<your-worker-name>.<your-subdomain>.workers.dev/sse
```

--------------------------------

### Connect MCP Inspector to Local Server

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-server-bearer-auth/static/README.md

Start the MCP Inspector and configure it to connect to a local MCP server via SSE. Ensure the correct URL and transport type are set.

```bash
npx @modelcontextprotocol/inspector
```

--------------------------------

### Run MCP Server Locally

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-google-oauth/README.md

Start the MCP server in local development mode. This makes the server accessible at http://localhost:8788 for testing.

```bash
wrangler dev
```

--------------------------------

### Summarization with Third-Party Providers

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Examples of performing summarization tasks using third-party AI providers via the AI Gateway. The `summarize` function is used with provider-specific creation functions.

```javascript
summarize() with createOpenAiSummarize
```

```javascript
summarize() with createAnthropicSummarize
```

```javascript
summarize() with createGeminiSummarize
```

--------------------------------

### Quick Start: Use Workers AI in a Cloudflare Worker

Source: https://github.com/cloudflare/ai/blob/main/packages/workers-ai-provider/README.md

Example of how to create a Cloudflare Worker that uses the Workers AI provider to stream text from a model.

```typescript
import { createWorkersAI } from "workers-ai-provider";
import { streamText } from "ai";

export default {
	async fetch(req: Request, env: { AI: Ai }) {
		const workersai = createWorkersAI({ binding: env.AI });

		const result = streamText({
			model: workersai("@cf/moonshotai/kimi-k2.7-code"),
			messages: [{ role: "user", content: "Write a haiku about Cloudflare" }],
		});

		return result.toTextStreamResponse();
	},
};
```

--------------------------------

### Build for Production

Source: https://github.com/cloudflare/ai/blob/main/demos/vision/README.md

Execute this command to create a production-ready build of the application.

```bash
npm run build
```

--------------------------------

### Structured Output Generation with Zod

Source: https://github.com/cloudflare/ai/blob/main/packages/workers-ai-provider/README.md

Generate structured output from a model using a Zod schema. This example shows how to define the expected output structure for a recipe and retrieve it.

```typescript
import { generateText, Output } from "ai";
import { z } from "zod";

const { output } = await generateText({
	model: workersai("@cf/moonshotai/kimi-k2.7-code"),
	prompt: "Recipe for spaghetti bolognese",
	output: Output.object({
		schema: z.object({
			name: z.string(),
			ingredients: z.array(z.object({ name: z.string(), amount: z.string() })),
			steps: z.array(z.string()),
		}),
	}),
});
```

--------------------------------

### Run Remote MCP Server Locally

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-server-bearer-auth/static/README.md

Start the remote MCP server locally using nx. This command is used to run the development server for testing.

```bash
npx nx dev remote-mcp-server-bearer-auth
```

--------------------------------

### API Response Format Example

Source: https://github.com/cloudflare/ai/blob/main/demos/structured-output/README.md

Defines the expected JSON structure for the server's response. This schema outlines the format for recipe details, including name, ingredients, and steps.

```json
{
	"recipe": {
		"name": "string",
		"ingredients": [{ "name": "string", "amount": "string" }],
		"steps": ["string"]
	}
}
```

--------------------------------

### Initialize Environment Configuration

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-consumer-todo-list/README.md

Copies template files to create local environment configuration files.

```bash
cp .env.template .env.local
```

```bash
cp .dev.vars.template .dev.vars
```

--------------------------------

### Configure Local Development Environment

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-google-oauth/README.md

Set up your local development environment by creating a `.dev.vars` file with your Google OAuth client ID and secret for local testing.

```bash
GOOGLE_CLIENT_ID=your_development_google_cloud_oauth_client_id
GOOGLE_CLIENT_SECRET=your_development_google_cloud_oauth_client_secret
```

--------------------------------

### Get Workflow Instance Status

Source: https://github.com/cloudflare/ai/blob/main/demos/orchestrator-workers/README.md

Fetches the status of an existing workflow instance by providing its ID in the URL. This is a GET request to the root endpoint with an ID parameter.

```bash
curl -X GET http://localhost:8787/{id}
```

--------------------------------

### API curl command

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling-stream/README.md

Example command to interact with the API endpoint.

```bash
curl -X POST \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is the weather in London?"}' \
  http://localhost:8787/
```

--------------------------------

### Configure local development environment variables

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-descope-auth/static/README.md

Create a .dev.vars file in the project root to store Descope credentials and the local server URL.

```bash
# .dev.vars
DESCOPE_PROJECT_ID="your_project_id"
DESCOPE_MANAGEMENT_KEY="your_management_key"
# For local development
SERVER_URL="http://localhost:8787"
```

--------------------------------

### GET /:id

Source: https://github.com/cloudflare/ai/blob/main/demos/evaluator-optimiser/README.md

Fetches the current status of a specific workflow instance.

```APIDOC
## GET /:id

### Description
Retrieves the status and progress details of an existing workflow instance using its unique ID.

### Method
GET

### Endpoint
/{id}

### Parameters
#### Path Parameters
- **id** (string) - Required - The unique identifier of the workflow instance.

### Response
#### Success Response (200)
- **status** (string) - The current status of the workflow instance.
```

--------------------------------

### Initialize AI Gateway with BYOK

Source: https://github.com/cloudflare/ai/blob/main/packages/ai-gateway-provider/README.md

Configure the gateway for unified billing or Bring Your Own Key scenarios.

```typescript
import { createAiGateway } from "ai-gateway-provider";
import { createOpenAI } from "ai-gateway-provider/providers/openai";
import { generateText } from "ai";

const aigateway = createAiGateway({
	accountId: "{CLOUDFLARE_ACCOUNT_ID}",
	gateway: "{GATEWAY_NAME}",
	apiKey: "{CF_AIG_TOKEN}",
});

const openai = createOpenAI();

const { text } = await generateText({
	model: aigateway(openai.chat("gpt-5.1")),
	prompt: "Write a vegetarian lasagna recipe for 4 people.",
});
```

--------------------------------

### GET /models_by_capability

Source: https://github.com/cloudflare/ai/blob/main/demos/model-scraper/README.md

Returns models that support specific capabilities, such as tool calling.

```APIDOC
## GET /models_by_capability

### Description
Returns models that support specific capabilities.

### Method
GET

### Endpoint
/models_by_capability

### Parameters
#### Query Parameters
- **capability** (string) - Required - The capability to filter models by (e.g., 'tools').

### Response
#### Success Response (200)
- **models** (array) - An array of models supporting the specified capability.
```

--------------------------------

### Configure local environment variables

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-auth0/todos-api/README.md

Create a .dev.vars file in the project root to store Auth0 configuration.

```text
AUTH0_DOMAIN=yourtenant.us.auth0.com
AUTH0_AUDIENCE=urn:todos-api
```

--------------------------------

### POST /

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager/README.md

Interact with the task manager agent by sending a prompt to perform task-related actions.

```APIDOC
## POST /

### Description
Interact with the task manager agent to perform operations such as adding, deleting, or listing tasks based on the provided prompt.

### Method
POST

### Endpoint
/

### Request Body
- **agentId** (string) - Required - The unique identifier for the agent.
- **prompt** (string) - Required - The user prompt describing the task action.

### Request Example
{
	"agentId": "your-agent-id",
	"prompt": "your-prompt"
}

### Response
#### Success Response (200)
- **message** (string) - Status message of the operation.
- **task** (object) - The task object affected by the operation.
  - **id** (string) - Unique task identifier.
  - **title** (string) - Task title.
  - **description** (string) - Task description.
  - **completed** (boolean) - Completion status.
  - **createdAt** (number) - Timestamp of creation.

#### Response Example
{
	"message": "Task created successfully",
	"task": {
		"id": "123",
		"title": "Example Task",
		"description": "Task description",
		"completed": false,
		"createdAt": 1672531200
	}
}
```

--------------------------------

### POST /

Source: https://github.com/cloudflare/ai/blob/main/demos/routing/README.md

Creates a new routing workflow instance to process a given prompt.

```APIDOC
## POST /

### Description
Creates a new routing workflow instance to process a given prompt.

### Method
POST

### Endpoint
/

### Request Body
- **prompt** (string) - Required - The prompt to be evaluated and processed.

### Request Example
{
	"prompt": "Your prompt here"
}

### Response
#### Success Response (200)
- **id** (string) - The unique identifier for the workflow instance.
- **details** (string) - Status details of the workflow.

#### Response Example
{
	"id": "instance-id",
	"details": "status-details"
}
```

--------------------------------

### Create KV Namespace

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-consumer-todo-list/README.md

Initializes a new KV namespace for the application data.

```bash
wrangler kv namespace create TODOS
```

--------------------------------

### Configure Local Development Environment Variables

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-descope-auth/README.md

Populate the .dev.vars file with your Descope credentials and a generated cookie encryption key. Ensure the encryption key is at least 32 bytes long.

```bash
# .dev.vars
DESCOPE_PROJECT_ID="your_project_id"
DESCOPE_MANAGEMENT_KEY="your_management_key"
COOKIE_ENCRYPTION_KEY="your_cookie_encryption_key"
```

--------------------------------

### Run Test Suite

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-scheduler/README.md

Executes the test suite for the application.

```bash
npx nx test agent-scheduler
```

--------------------------------

### POST /

Source: https://github.com/cloudflare/ai/blob/main/demos/prompt-chaining/README.md

Triggers a new workflow instance for generating technical documentation.

```APIDOC
## POST /

### Description
Triggers a new workflow instance to begin the automated documentation generation process.

### Method
POST

### Endpoint
/

### Parameters
#### Request Body
- **prompt** (string) - Required - The input prompt used to initiate the documentation generation workflow.

### Request Example
{
  "prompt": "Your prompt here"
}

### Response
#### Success Response (200)
- **id** (string) - The unique identifier for the workflow instance.
- **details** (object) - Details regarding the initialized workflow instance.

#### Response Example
{
  "id": "workflow-123",
  "details": { "status": "initiated" }
}
```

--------------------------------

### GET /models_by_task

Source: https://github.com/cloudflare/ai/blob/main/demos/model-scraper/README.md

Returns an array of models whose task.id matches the provided task parameter.

```APIDOC
## GET /models_by_task

### Description
Returns an array of models whose task.id matches the provided task parameter.

### Method
GET

### Endpoint
/models_by_task

### Parameters
#### Query Parameters
- **task** (string) - Required - The task ID to filter models by.

### Response
#### Success Response (200)
- **models** (array) - An array of models matching the task ID.
```

--------------------------------

### Run Test Suite

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager/README.md

Executes the project test suite.

```bash
npx nx test agent-task-manager
```

--------------------------------

### Interact with API via cURL

Source: https://github.com/cloudflare/ai/blob/main/demos/ui-worker/README.md

Example request to the /api endpoint for triggering text generation tasks.

```bash
curl -X POST \
  http://localhost:8787/api \
  -H 'Content-Type: application/json' \
  -d '{"prompt": "Your task description here"}'
```

--------------------------------

### NPM scripts for project management

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling-stream/README.md

Commonly used commands for deploying, linting, testing, and type-checking the project.

```bash
npx nx deploy tool-calling-stream
```

```bash
npx nx dev tool-calling-stream
```

```bash
npx nx lint tool-calling-stream
```

```bash
npx nx start tool-calling-stream
```

```bash
npx nx test tool-calling-stream
```

```bash
npx nx test:ci tool-calling-stream
```

```bash
npx nx type-check tool-calling-stream
```

--------------------------------

### NPM Scripts for Project Management

Source: https://github.com/cloudflare/ai/blob/main/demos/routing/README.md

Standard commands for deploying, linting, testing, and type-checking the routing application.

```bash
npx nx deploy routing
```

```bash
npx nx dev routing
```

```bash
npx nx lint routing
```

```bash
npx nx start routing
```

```bash
npx nx test routing
```

```bash
npx nx test:ci routing
```

```bash
npx nx type-check routing
```

--------------------------------

### GET /:id - Fetch Workflow Status

Source: https://github.com/cloudflare/ai/blob/main/demos/parallelisation/README.md

Fetches the status of an existing workflow instance using its unique identifier.

```APIDOC
## GET /:id

### Description
Fetches the status of an existing workflow instance by its ID.

### Method
GET

### Endpoint
/:id

### Parameters
#### Path Parameters
- **id** (string) - Required - The unique identifier of the workflow instance.

### Response
#### Success Response (200)
- **status** (string) - The current status of the workflow instance.

#### Response Example
```json
{
  "status": "processing"
}
```
```

--------------------------------

### API Response for Workflow Status

Source: https://github.com/cloudflare/ai/blob/main/demos/orchestrator-workers/README.md

Example JSON response when fetching the status of a workflow instance. It provides the current status of the workflow.

```json
{
	"status": "current-status"
}
```

--------------------------------

### Create KV Namespace

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-logto/README.md

Command to initialize a KV namespace for storing OAuth data.

```bash
npx wrangler kv namespace create "OAUTH_KV"
```

--------------------------------

### Curl Command for Confirmation

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager-human-in-the-loop/README.md

Example cURL command to confirm or deny an action via the /confirmations/:confirmationId endpoint. Replace placeholders.

```bash
curl -X POST http://localhost:8787/confirmations/{confirmationId} -H "Content-Type: application/json" -d '{"agentId": "your-agent-id", "confirm": true}'
```

--------------------------------

### Deploy to Cloudflare

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server/README.md

Commands to create a KV namespace and deploy the worker application.

```bash
npx wrangler kv namespace create OAUTH_KV
# Follow the guidance to add the kv namespace ID to wrangler.jsonc
npm run deploy
```

--------------------------------

### Curl Command for Task Query

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager-human-in-the-loop/README.md

Example cURL command to send a query to the /query endpoint. Replace placeholders with actual values.

```bash
curl -X POST http://localhost:8787/query -H "Content-Type: application/json" -d '{"agentId": "your-agent-id", "prompt": "your-prompt"}'
```

--------------------------------

### Initialize AI Gateway with API Key

Source: https://github.com/cloudflare/ai/blob/main/packages/ai-gateway-provider/README.md

Configure the gateway using Cloudflare account credentials and an optional API key.

```typescript
import { createAiGateway } from "ai-gateway-provider";
import { createOpenAI } from "ai-gateway-provider/providers/openai";
import { generateText } from "ai";

const aigateway = createAiGateway({
	accountId: "{CLOUDFLARE_ACCOUNT_ID}",
	gateway: "{GATEWAY_NAME}",
	apiKey: "{CF_AIG_TOKEN}", // If your AI Gateway has authentication enabled
});

const openai = createOpenAI({ apiKey: "{OPENAI_API_KEY}" });

const { text } = await generateText({
	model: aigateway(openai.chat("gpt-5.1")),
	prompt: "Write a vegetarian lasagna recipe for 4 people.",
});
```

--------------------------------

### Workers AI Direct Chat

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Example of creating a chat interface using Workers AI directly. Requires an AI binding configured in your environment.

```javascript
createWorkersAiChat(model, { binding: env.AI })
```

--------------------------------

### API Response for Triggering Workflow

Source: https://github.com/cloudflare/ai/blob/main/demos/orchestrator-workers/README.md

Example JSON response when a new workflow instance is successfully triggered. It includes an ID for tracking and status details.

```json
{
	"id": "instance-id",
	"details": "status-details"
}
```

--------------------------------

### Deploy Application

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager/README.md

Deploys the application using Wrangler.

```bash
npx nx deploy agent-task-manager
```

--------------------------------

### Run Test Suite

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling-stream-traditional/README.md

Runs the project's test suite using Vitest. This is useful for verifying functionality.

```bash
npx nx test tool-calling-stream-traditional
```

--------------------------------

### Lint Python code with Ruff

Source: https://github.com/cloudflare/ai/blob/main/demos/python-workers-mcp/README.md

Run this command to lint your Python code using Ruff, which checks for style guide violations and potential errors.

```console
uv ruff check .
```

--------------------------------

### Create Remote MCP Server with GitHub OAuth Template

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-github-oauth/README.md

Use this command to scaffold a new remote MCP server project pre-configured with GitHub OAuth integration.

```bash
npm create cloudflare@latest -- my-mcp-server --template=cloudflare/ai/demos/remote-mcp-github-oauth
```

--------------------------------

### Update Stytch RBAC Policy

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-b2b-okr-manager/README.md

Use this command to update your Stytch RBAC policy with your workspace management credentials and project ID. Replace example credentials with your own.

```bash
npm run update-policy -- --key-id "workspace-key-prod-4881b817-6336-410a-a953-6eceabaf5xc9" --secret "6ZcNGH7v9Oxxxxxxxxxx" --project-id "project-test-6c20cd16-73d5-44f7-852c-9a7e7b2ccf62"
```

--------------------------------

### Configure Local Environment Variables (.env.local)

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-b2b-okr-manager/README.md

Set your Stytch public token and project domain for local development.

```env
VITE_STYTCH_PUBLIC_TOKEN=public-token-test-abc123-abcde-1234-0987-0000-abcd1234
VITE_STYTCH_DOMAIN=https://project-domain.customers.stytch.dev
```

--------------------------------

### Deploy to Cloudflare

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-autorag/static/README.md

Create a KV namespace, update wrangler.jsonc with the namespace ID, and deploy the worker.

```bash
npx wrangler@latest kv namespace create remote-mcp-server-oauth-kv
npm run deploy
```

--------------------------------

### Image Generation with Third-Party Providers

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Examples of generating images using third-party AI providers through the AI Gateway. The `generateImage` function is used with provider-specific creation functions.

```javascript
generateImage() with createOpenAiImage
```

```javascript
generateImage() with createGeminiImage
```

```javascript
generateImage() with createGrokImage
```

--------------------------------

### POST /

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling-stream/README.md

Processes a user prompt and returns a streaming response from the AI model, which may involve tool execution.

```APIDOC
## POST /

### Description
Processes a user prompt and returns a streaming response. The AI model identifies tasks and interacts with external tools as needed.

### Method
POST

### Endpoint
/

### Request Body
- **prompt** (string) - Required - The user prompt to be processed by the AI model.

### Request Example
{
	"prompt": "What is the weather in London?"
}

### Response
#### Success Response (200)
- **response** (stream) - A streaming response containing the AI's output and tool execution results.

#### Response Example
{
	"response": "The weather in London is currently 15°C with light rain."
}
```

--------------------------------

### Third-Party Chat Provider Integration

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Examples of creating chat interfaces for third-party AI providers via the AI Gateway. These functions abstract the provider-specific API calls.

```javascript
createOpenAiChat
```

```javascript
createAnthropicChat
```

```javascript
createGeminiChat
```

```javascript
createGrokChat
```

--------------------------------

### Workers AI via AI Gateway Chat

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Example of creating a chat interface using Workers AI through the AI Gateway. Requires the AI Gateway ID.

```javascript
createWorkersAiChat(model, { binding: env.AI.gateway(id) })
```

--------------------------------

### System Architecture Diagram

Source: https://github.com/cloudflare/ai/blob/main/demos/agent-task-manager/README.md

Mermaid diagram illustrating the flow from client to task storage.

```mermaid
graph TD;
    A[Client] --> B[API Server];
    B --> C[TaskManagerAgent];
    C --> D[Task Storage];
```

--------------------------------

### System Diagram for Tool Calling

Source: https://github.com/cloudflare/ai/blob/main/demos/tool-calling/README.md

Illustrates the data flow from client to server for processing weather-related prompts and fetching data.

```mermaid
graph TD;
    A[Client] -->|POST Request| B[DevServerTestHelper]
    B -->|Start Server| C[Weather Worker]
    C -->|Process Prompt| D[Fetch Weather Data]
    D -->|Return Response| A
```

--------------------------------

### Generate Text with Workers AI

Source: https://github.com/cloudflare/ai/blob/main/packages/workers-ai-provider/README.md

Use the `generateText` function to get a text response from a Workers AI model. Ensure the model is correctly specified using the `workersai` helper.

```typescript
import { generateText } from "ai";

const { text } = await generateText({
	model: workersai("@cf/moonshotai/kimi-k2.7-code"),
	prompt: "Explain Workers AI in one paragraph",
});
```

--------------------------------

### Configure Environment Variables

Source: https://github.com/cloudflare/ai/blob/main/examples/tanstack-ai/README.md

Fill in your Cloudflare credentials into the .dev.vars file. These variables are used for authentication and deployment.

```bash
CLOUDFLARE_ACCOUNT_ID=your-cloudflare-account-id
CLOUDFLARE_AI_GATEWAY_ID=your-ai-gateway-id
CLOUDFLARE_API_TOKEN=your-cloudflare-api-token
```

--------------------------------

### POST /

Source: https://github.com/cloudflare/ai/blob/main/demos/evaluator-optimiser/README.md

Triggers a new workflow instance for text optimization.

```APIDOC
## POST /

### Description
Triggers a new workflow instance to begin the iterative drafting and optimization process.

### Method
POST

### Endpoint
/

### Parameters
#### Request Body
- **prompt** (string) - Required - The initial text prompt to be processed by the workflow.

### Request Example
{
  "prompt": "Your prompt here"
}

### Response
#### Success Response (200)
- **instance_id** (string) - The unique identifier for the created workflow instance.
- **status** (string) - The current status of the workflow.
```

--------------------------------

### Tool Calling with Zod Schema

Source: https://github.com/cloudflare/ai/blob/main/packages/workers-ai-provider/README.md

Utilize tool calling functionality with a Zod schema for defining tool inputs. This example demonstrates how to define a 'getWeather' tool and use it within the generateText function.

```typescript
import { generateText, stepCountIs } from "ai";
import { z } from "zod";

const { text } = await generateText({
	model: workersai("@cf/moonshotai/kimi-k2.7-code"),
	prompt: "What's the weather in London?",
	tools: {
		getWeather: {
			description: "Get the current weather for a city",
			inputSchema: z.object({ city: z.string() }),
			execute: async ({ city }) => ({ city, temperature: 18, condition: "Cloudy" }),
		},
	},
	stopWhen: stepCountIs(2),
});
```

--------------------------------

### Configure Local Environment Variables (.dev.vars)

Source: https://github.com/cloudflare/ai/blob/main/demos/mcp-stytch-b2b-okr-manager/README.md

Set your Stytch project ID, secret, and domain for local development. Ensure these are kept confidential.

```env
STYTCH_PROJECT_ID=project-test-6c20cd16-73d5-44f7-852c-9a7e7b2ccf62
STYTCH_PROJECT_SECRET=secret-test-.....
STYTCH_DOMAIN=https://project-domain.customers.stytch.dev
```

--------------------------------

### Chat with Reasoning Controls

Source: https://github.com/cloudflare/ai/blob/main/packages/tanstack-ai/README.md

Configure chat models to use reasoning capabilities like 'reasoning_effort' and 'chat_template_kwargs'. This example uses a GLM-4.7-flash model and sets reasoning effort to 'low' and disables thinking.

```typescript
import { chat } from "@tanstack/ai";
import { createWorkersAiChat } from "@cloudflare/tanstack-ai";

const adapter = createWorkersAiChat("@cf/zai-org/glm-4.7-flash", {
	binding: env.AI,
});

const response = chat({
	adapter,
	stream: true,
	messages: [{ role: "user", content: "Summarize in one sentence." }],
	modelOptions: {
		// "low" | "medium" | "high" | null — null disables reasoning.
		reasoning_effort: "low",
		// Toggle thinking on models that expose template kwargs (GLM, Kimi).
		chat_template_kwargs: { enable_thinking: false },
	},
});
```

--------------------------------

### Deploy to Cloudflare

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server-autorag/README.md

Steps to deploy the application to Cloudflare Workers, including creating a KV namespace and running the deployment command.

```bash
npx wrangler kv namespace create OAUTH_KV

npm run deploy
```

--------------------------------

### Configure Claude Desktop

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-cf-access-self-hosted/README.md

Add the self-hosted MCP server configuration to Claude Desktop's settings.

```json
{
	"mcpServers": {
		"access-self-hosted": {
			"type": "http",
			"url": "https://mcp-access-self-hosted.<your-subdomain>.workers.dev/mcp"
		}
	}
}
```

--------------------------------

### Configure Claude Desktop for local MCP server

Source: https://github.com/cloudflare/ai/blob/main/demos/remote-mcp-server/static/README.md

JSON configuration for the Claude Desktop app to connect to a local MCP server via a proxy.

```json
{
	"mcpServers": {
		"math": {
			"command": "npx",
			"args": ["mcp-remote", "http://localhost:8787/sse"]
		}
	}
}
```