Eko Framework

https://fellou.ai/eko/docs/
Admin
Eko Framework enables developers to build production-ready agentic workflows using natural language,...
Tokens:24,202
Snippets:217
Trust Score:8.3
Update:1 year ago
Show doc for...
Tokens:
Raw
### Run Fellou AI Eko Node.js Script

Source: https://fellou.ai/eko/docs/getting-started/quickstart

Executes the Fellou AI Eko Node.js script after setting up the project and installing dependencies. This command starts the workflow defined in the script.

```bash
npm run start
```

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

### Install Node.js Dependencies for Fellou AI Eko

Source: https://fellou.ai/eko/docs/getting-started/quickstart

Installs the necessary dependencies for running a Fellou AI Eko Node.js script. This typically includes project setup and package management.

```bash
npm install
```

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

### Configure Eko with a single BrowserAgent

Source: https://fellou.ai/eko/docs/getting-started/configuration

Sets up a single BrowserAgent for the Eko workflow. This example shows how to define an agent with its name, description, toolkit, and the models it can use.

```python
eko_config = EkoConfig(
    agents=[
        AgentConfig(
            name="BrowserAgent",
            description="A web browser agent that can navigate and interact with web pages.",
            toolkit=["browser"],
            models=["gpt-4o", "claude-3-5-sonnet-20240620"],
        )
    ]
)
```

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

### Eko Quickstart Setup and Execution

Source: https://fellou.ai/eko/docs/api/index

Demonstrates how to set up Eko by defining LLMs and agents, then running a task. It requires API keys for LLM providers and uses `ts-node` for execution.

```typescript
const llms: LLMs = {
  default: {
    provider: "anthropic",
    model: "claude-3-5-sonnet-20241022",
    apiKey: claudeApiKey || "your-api-key",
    config: { baseURL: claudeBaseURL },
  },
  openai: {
    provider: "openai",
    model: "gpt-4o-mini",
    apiKey: openaiApiKey || "your-api-key",
    config: { baseURL: openaiBaseURL },
  },
};
let agents: Agent[] = [new ChatAgent(), new BrowserAgent()];
let eko = new Eko({ llms, agents });
let result = await eko.run("Search for the latest news about Musk");
```

```bash
pnpm install @eko-ai/eko
npx ts-node quickstart.ts
```

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

### Eko API Constructor

Source: https://fellou.ai/eko/docs/api/classes/Eko

Initializes a new Eko instance with the provided configuration. This is the primary way to start using the Eko API.

```typescript
new Eko(config: EkoConfig): Eko
```

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

### Write Fellou AI Eko Node.js Script (index.ts)

Source: https://fellou.ai/eko/docs/getting-started/quickstart

Demonstrates the basic structure of a TypeScript script for Fellou AI Eko, including defining imports and a main execution function.

```typescript
import { Eko } from "@ekoai/core";

async function main() {
  const eko = new Eko({
    // You can use one of the following environments:
    // OPENAI_API_KEY: process.env.OPENAI_API_KEY,
    // CLAUDE_API_KEY: process.env.CLAUDE_API_KEY,
  });

  const res = await eko.run(
    "Open Twitter, search for \"Fellou AI\" and follow"
  );

  console.log(res);
}

main();
```

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

### Callback Usage Example with Logging and Confirmation

Source: https://fellou.ai/eko/docs/architecture/callback-system

An example demonstrating how to use callbacks in Eko, specifically implementing `onMessage` for logging workflow and LLM responses, and `onHumanConfirm` for user confirmation of operations.

```python
from typing import Any, Dict

def on_message(message: Dict[str, Any]) -> None:
    # Log workflow, LLM responses, and tool invocation parameters
    print(f"Received message: {message}")

def on_human_confirm(context: Dict[str, Any], prompt: str) -> bool:
    # Allow users to confirm certain operations
    print(f"Confirmation required: {prompt}")
    user_confirmation = input("Do you confirm? (yes/no): ")
    return user_confirmation.lower() == 'yes'

# Example of how these might be registered or used within a workflow
# Assume 'workflow_instance' is an object that accepts callback handlers
# workflow_instance.register_callback('onMessage', on_message)
# workflow_instance.register_callback('onHumanConfirm', on_human_confirm)

```

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

### Configure Eko LLMs with Claude 3.5 and GPT-4o

Source: https://fellou.ai/eko/docs/getting-started/configuration

Configures two large language models for Eko: Claude 3.5 as the default and GPT-4o. This demonstrates setting provider, model name, API key, base URL, TopK, and TopP for different models.

```python
eko_config = EkoConfig(
    llms=[
        LLMConfig(
            provider="anthropic",
            model_name="claude-3-5-sonnet-20240620",
            api_key="YOUR_ANTHROPIC_API_KEY",
            base_url="https://api.anthropic.com/v1",
            top_k=50,
            top_p=0.9,
        ),
        LLMConfig(
            provider="openai",
            model_name="gpt-4o",
            api_key="YOUR_OPENAI_API_KEY",
            base_url="https://api.openai.com/v1",
            top_k=50,
            top_p=0.9,
        ),
    ]
)
```

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

### Example Eko Workflow XML

Source: https://fellou.ai/eko/docs/architecture

An example of an Eko Workflow defined in XML, illustrating the structure for task execution. It includes tags for workflow name, thought process, agents, and nodes representing subtasks.

```xml
<workflow>
  <name>Open Twitter and Follow</name>
  <thought>I need to open Twitter, search for Fellou AI, and follow the account.</thought>
  <agents>
    <agent name="Browser"/>
  </agents>
  <nodes>
    <node name="Open Twitter" agent="Browser" tool="open_website" params="twitter.com"/>
    <node name="Search Fellou AI" agent="Browser" tool="search" params="Fellou AI"/>
    <node name="Follow" agent="Browser" tool="click" params="Follow Button"/>
  </nodes>
</workflow>
```

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

### Configure Eko Callbacks with StreamCallback and HumanCallback

Source: https://fellou.ai/eko/docs/getting-started/configuration

Sets up user-defined callback functions for Eko, including StreamCallback for progress updates and HumanCallback for handling user input. This allows for real-time feedback and interactive workflows.

```python
from eko.callbacks import StreamCallback, HumanCallback

my_stream_callback = StreamCallback()
my_human_callback = HumanCallback()

eko_config = EkoConfig(
    callback=[
        my_stream_callback,
        my_human_callback,
    ]
)
```

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

### Dynamically Expand Agent Tool with MCP

Source: https://fellou.ai/eko/docs/agents/mcp-tools

Provides an example of how to dynamically expand an agent's tools using the Model Context Protocol (MCP). This assumes an MCP service is already available and accessible.

```typescript
async function expandAgentTools(mcpClient: IMcpClient) {
    // Logic to load and register tools using the MCP client
    console.log("Dynamically expanding agent tools...");
    // Example: const toolDefinitions = await mcpClient.getTools();
    // agent.addTools(toolDefinitions);
}
```

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

### Get BaseChatAgent PlanDescription

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Retrieves the plan description for the BaseChatAgent instance. This accessor returns a string or undefined.

```typescript
get PlanDescription(): undefined | string
```

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

### Configure Eko Planner LLMs

Source: https://fellou.ai/eko/docs/getting-started/configuration

Specifies the large language model to be used by the Eko Planner. It's recommended to use a high-performance model for planning complex workflows.

```python
eko_config = EkoConfig(
    plan_llms=[
        LLMConfig(
            provider="openai",
            model_name="gpt-4o",
            api_key="YOUR_OPENAI_API_KEY",
            base_url="https://api.openai.com/v1",
        )
    ]
)
```

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

### Get BaseChatAgent Description

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Retrieves the description of the BaseChatAgent instance. This accessor returns a string representing the agent's description.

```typescript
get Description(): string
```

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

### Get BaseChatAgent Tools

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Retrieves the list of tools associated with the BaseChatAgent instance. This accessor returns an array of Tool objects.

```typescript
get Tools(): Tool[]
```

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

### Browser Agent Element Interaction Tools

Source: https://fellou.ai/eko/docs/agents/browser-agent

Enables interaction with web page elements using unique indexes. Supports typing text, clicking, hovering, scrolling to elements, and selecting options from dropdowns. Includes methods to get dropdown options.

```JavaScript
input_text(index: number, text: string, enter?: boolean)
click_element(index: number, num_clicks?: number, button?: "left" | "right" | "middle")
hover_to_element(index: number)
scroll_to_element(index: number)
select_option(index: number, option_value: string)
get_select_options(index: number)
```

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

### Browser Agent Content Extraction Tools

Source: https://fellou.ai/eko/docs/agents/browser-agent

Facilitates the extraction of page content and structural information. Includes methods to get the full page content and to capture a screenshot along with HTML data of interactive elements.

```JavaScript
extract_page_content(variable_name?: string)
screenshot_and_html()
```

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

### SimpleSseMcpClient Constructor

Source: https://fellou.ai/eko/docs/api/classes/SimpleSseMcpClient

Initializes a new instance of the SimpleSseMcpClient class.

```APIDOC
## Constructor SimpleSseMcpClient

### Description
Initializes a new instance of the SimpleSseMcpClient class.

### Parameters
#### Parameters
- **sseServerUrl** (string) - Required - The URL of the SSE server.
- **clientName** (string) - Optional - The name of the client. Defaults to "EkoMcpClient".
```

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

### BaseBrowserScreenAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseBrowserScreenAgent

Initializes a new instance of the BaseBrowserScreenAgent class.

```APIDOC
## Constructors

### constructor

* `new BaseBrowserScreenAgent(
llms?: string[],
ext_tools?: Tool[],
mcpClient?: IMcpClient
): BaseBrowserScreenAgent

#### Parameters

* `Optional` llms: string[]
* `Optional` ext_tools: Tool[]
* `Optional` mcpClient: IMcpClient

#### Returns

BaseBrowserScreenAgent
```

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

### Get BaseChatAgent Name

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Retrieves the name of the BaseChatAgent instance. This accessor returns a string representing the agent's name.

```typescript
get Name(): string
```

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

### SimpleSseMcpClient.connect

Source: https://fellou.ai/eko/docs/api/classes/SimpleSseMcpClient

Establishes a connection to the SSE server.

```APIDOC
## Methods connect

### Description
Establishes a connection to the SSE server.

### Method
`POST`

### Endpoint
`/connect`

### Returns
- **Promise<void>** - A promise that resolves when the connection is established.
```

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

### BaseBrowserScreenAgent Methods

Source: https://fellou.ai/eko/docs/api/classes/BaseBrowserScreenAgent

Contains methods for running the agent with or without context.

```APIDOC
## Methods

### run

* `run(context: Context, agentChain: AgentChain): Promise<string>`

#### Parameters

* context: Context
* agentChain: AgentChain

#### Returns

Promise<string>

### runWithContext

* `runWithContext(
agentContext: AgentContext,
mcpClient?: IMcpClient,
maxReactNum?: number = 100
): Promise<string>

#### Parameters

* agentContext: AgentContext
* `Optional` mcpClient: IMcpClient
* maxReactNum: number = 100

#### Returns

Promise<string>
```

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

### Get BaseChatAgent McpClient

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Retrieves the MCP client associated with the BaseChatAgent instance. This accessor returns either an IMcpClient object or undefined.

```typescript
get McpClient(): undefined | IMcpClient
```

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

### Instantiate BaseBrowserLabelsAgent

Source: https://fellou.ai/eko/docs/api/classes/BaseBrowserLabelsAgent

This code snippet demonstrates how to create an instance of the BaseBrowserLabelsAgent class. It shows the optional parameters that can be passed during initialization, including language models, external tools, and an MCP client.

```typescript
new BaseBrowserLabelsAgent(
  llms?: string[],
  ext_tools?: Tool[],
  mcpClient?: IMcpClient,
): BaseBrowserLabelsAgent
```

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

### BaseShellAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseShellAgent

Initializes a new instance of the BaseShellAgent class. Supports optional LLMs, external tools, MCP client, and plan description.

```APIDOC
## Constructors

### constructor

```typescript
new BaseShellAgent(
  llms?: string[],
  ext_tools?: Tool[],
  mcpClient?: IMcpClient,
  planDescription?: string
): BaseShellAgent
```

#### Parameters

* `Optional` llms: string[] - An array of language model identifiers.
* `Optional` ext_tools: Tool[] - An array of external tools to be used by the agent.
* `Optional` mcpClient: IMcpClient - An instance of the MCP client.
* `Optional` planDescription: string - A description of the agent's plan.
```

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

### SimpleSseMcpClient.listTools

Source: https://fellou.ai/eko/docs/api/classes/SimpleSseMcpClient

Retrieves a list of available tools from the SSE server.

```APIDOC
## Methods listTools

### Description
Retrieves a list of available tools from the SSE server.

### Method
`POST`

### Endpoint
`/listTools`

### Parameters
#### Request Body
- **param** (McpListToolParam) - Required - Parameters for listing tools.

### Returns
- **Promise<McpListToolResult>** - A promise that resolves with the list of tools.
```

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

### SimpleSseMcpClient Constructor

Source: https://fellou.ai/eko/docs/api/classes/SimpleSseMcpClient

Initializes a new instance of the SimpleSseMcpClient. It requires the SSE server URL and optionally accepts a client name.

```typescript
new SimpleSseMcpClient(
  sseServerUrl: string,
  clientName?: string
): SimpleSseMcpClient
```

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

### BaseShellAgent Methods

Source: https://fellou.ai/eko/docs/api/classes/BaseShellAgent

Details the core methods of the BaseShellAgent for executing tasks, including `run` and `runWithContext`.

```APIDOC
## Methods

### run

* **run(context: Context, agentChain: AgentChain): Promise<string>** - Executes the agent's process with the given context and agent chain.

#### Parameters

* `context`: Context - The execution context for the agent.
* `agentChain`: AgentChain - The chain of agents to be executed.

### runWithContext

* **runWithContext(agentContext: AgentContext, mcpClient?: IMcpClient, maxReactNum?: number): Promise<string>** - Executes the agent with a specific context, optionally providing an MCP client and a maximum reaction number.

#### Parameters

* `agentContext`: AgentContext - The context object for the agent's operation.
* `Optional` mcpClient: IMcpClient - An optional MCP client instance.
* `maxReactNum`: number = 100 - The maximum number of reactions allowed during execution. Defaults to 100.
```

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

### EKO API - Constructor

Source: https://fellou.ai/eko/docs/api/classes/Eko

Initializes a new instance of the Eko class with the provided configuration.

```APIDOC
## new Eko(config: EkoConfig): Eko

### Description
Initializes a new instance of the Eko class.

### Parameters
* **config** (EkoConfig) - The configuration object for the Eko instance.

### Returns
Eko - An instance of the Eko class.
```

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

### BaseComputerAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseComputerAgent

Initializes a new instance of the BaseComputerAgent. Allows for optional configuration of Large Language Models (LLMs), external tools, an MCP client, and keyboard keys.

```typescript
new BaseComputerAgent(
  llms?: string[],
  ext_tools?: Tool[],
  mcpClient?: IMcpClient,
  keyboardKeys?: string[],
): BaseComputerAgent
```

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

### BaseBrowserScreenAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseBrowserScreenAgent

Initializes a new instance of the BaseBrowserScreenAgent class. It accepts optional arrays of LLM strings, tool objects, and an MCP client instance.

```typescript
new BaseBrowserScreenAgent(
llms?: string[],
ext_tools?: Tool[],
mcpClient?: IMcpClient,
): BaseBrowserScreenAgent
```

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

### BaseBrowserLabelsAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseBrowserLabelsAgent

Initializes a new instance of the BaseBrowserLabelsAgent class.

```APIDOC
## Constructor BaseBrowserLabelsAgent

### Description
Initializes a new instance of the BaseBrowserLabelsAgent class.

### Method
constructor

### Parameters
* `Optional` **llms**: string[] - Optional array of language models.
* `Optional` **ext_tools**: Tool[] - Optional array of external tools.
* `Optional` **mcpClient**: IMcpClient - Optional MCP client instance.

### Returns
BaseBrowserLabelsAgent - An instance of the BaseBrowserLabelsAgent class.
```

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

### BaseFileAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseFileAgent

Initializes a new instance of the BaseFileAgent class. It allows setting the working path, language models, external tools, and MCP client.

```APIDOC
## Constructors

### constructor

Creates an instance of `BaseFileAgent`.

#### Parameters

* `work_path` (string, Optional): The working directory path for the agent.
* `llms` (string[], Optional): An array of language model identifiers.
* `ext_tools` (Tool[], Optional): An array of external tools to be used by the agent.
* `mcpClient` (IMcpClient, Optional): An instance of the MCP client.
* `planDescription` (string, Optional): A description of the agent's plan.

#### Returns

* `BaseFileAgent`: A new instance of the BaseFileAgent class.
```

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

### BaseComputerAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseComputerAgent

Initializes a new instance of the BaseComputerAgent class with optional parameters.

```APIDOC
## Constructors

### constructor
  * new BaseComputerAgent(
llms?: string[],
ext_tools?: Tool[],
mcpClient?: IMcpClient,
keyboardKeys?: string[]
): BaseComputerAgent
#### Parameters
    * `Optional`llms: string[]
    * `Optional`ext_tools: Tool[]
    * `Optional`mcpClient: IMcpClient
    * `Optional`keyboardKeys: string[]
#### Returns BaseComputerAgent
```

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

### VariableStorageTool Constructor

Source: https://fellou.ai/eko/docs/api/classes/VariableStorageTool

Initializes a new instance of the VariableStorageTool.

```APIDOC
## Constructors
### constructor
  * new VariableStorageTool(): VariableStorageTool
#### Returns VariableStorageTool
```

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

### BaseBrowserLabelsAgent Methods

Source: https://fellou.ai/eko/docs/api/classes/BaseBrowserLabelsAgent

Provides methods for running the agent with or without context.

```APIDOC
## Method run BaseBrowserLabelsAgent

### Description
Executes the agent's run logic with the provided context and agent chain.

### Method
run(context: Context, agentChain: AgentChain)

### Parameters
* **context**: Context - The context for the agent execution.
* **agentChain**: AgentChain - The agent chain to be used.

### Returns
Promise<string> - A promise that resolves to a string representing the result of the execution.
```

```APIDOC
## Method runWithContext BaseBrowserLabelsAgent

### Description
Executes the agent's logic with the provided agent context.

### Method
runWithContext(agentContext: AgentContext, mcpClient?: IMcpClient, maxReactNum?: number)

### Parameters
* **agentContext**: AgentContext - The context for the agent execution.
* `Optional` **mcpClient**: IMcpClient - An optional MCP client instance.
* **maxReactNum**: number = 100 - The maximum number of reactions allowed (defaults to 100).

### Returns
Promise<string> - A promise that resolves to a string representing the result of the execution.
```

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

### Browser Agent - Navigation Tools

Source: https://fellou.ai/eko/docs/agents/browser-agent

Tools for navigating web pages and managing browser tabs.

```APIDOC
## POST /browser_agent/navigate_to

### Description
Navigates the browser to a specified URL.

### Method
POST

### Endpoint
/browser_agent/navigate_to

### Parameters
#### Request Body
- **url** (string) - Required - The URL to navigate to.

### Request Example
```json
{
  "url": "https://www.example.com"
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the navigation.

#### Response Example
```json
{
  "status": "navigated successfully"
}
```

## POST /browser_agent/go_back

### Description
Navigates the browser back to the previous page in the history.

### Method
POST

### Endpoint
/browser_agent/go_back

### Parameters
No parameters required.

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the back navigation.

#### Response Example
```json
{
  "status": "navigated back successfully"
}
```

## POST /browser_agent/switch_tab

### Description
Switches the active browser tab to a specified tab index.

### Method
POST

### Endpoint
/browser_agent/switch_tab

### Parameters
#### Request Body
- **index** (number) - Required - The index of the tab to switch to.

### Request Example
```json
{
  "index": 1
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the tab switch.

#### Response Example
```json
{
  "status": "switched to tab 1"
}
```

## GET /browser_agent/get_all_tabs

### Description
Retrieves a list of all currently open browser tabs.

### Method
GET

### Endpoint
/browser_agent/get_all_tabs

### Parameters
No parameters required.

### Response
#### Success Response (200)
- **tabs** (array) - A list of all open tabs, each containing its URL and title.

#### Response Example
```json
{
  "tabs": [
    { "url": "https://www.example.com", "title": "Example Domain" },
    { "url": "https://www.google.com", "title": "Google" }
  ]
}
```
```

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

### WatchTriggerTool Constructor

Source: https://fellou.ai/eko/docs/api/classes/WatchTriggerTool

Initializes a new instance of the WatchTriggerTool.

```APIDOC
## new WatchTriggerTool()

### Description
Initializes a new instance of the WatchTriggerTool class.

### Method
CONSTRUCTOR

### Endpoint
N/A

### Returns
WatchTriggerTool - A new instance of the WatchTriggerTool.

### Request Example
```typescript
const triggerTool = new WatchTriggerTool();
```

### Response Example
```json
{
  "message": "WatchTriggerTool instance created successfully"
}
```
```

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

### BaseShellAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseShellAgent

Initializes a new instance of the BaseShellAgent class. It accepts optional parameters for language models, external tools, an MCP client, and a plan description.

```typescript
new BaseShellAgent(
  llms?: string[],
  ext_tools?: Tool[],
  mcpClient?: IMcpClient,
  planDescription?: string,
): BaseShellAgent
```

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

### Initialize SimpleSseMcpClient

Source: https://fellou.ai/eko/docs/agents/mcp-tools

Demonstrates the initialization of the `SimpleSseMcpClient` class, which is an implementation of the MCP interface using the Server-Sent Events (SSE) protocol. This client is used to connect to MCP services.

```typescript
const mcpClient = new SimpleSseMcpClient("http://your-mcp-service.com");
```

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

### Eko Planning and Execution

Source: https://fellou.ai/eko/docs/architecture

Demonstrates the core methods for Eko's execution model. `Eko.generate` handles task breakdown into workflows, while `Eko.execute` or `Eko.run` drives the agent-based execution.

```python
from ello import Eko

# Planning Phase
workflow = Eko.generate("Open https://fellou.ai and generate a summary.md")

# Execution Phase
result = Eko.execute(workflow)
# or
# result = Eko.run(workflow)
```

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

### AgentChain Methods

Source: https://fellou.ai/eko/docs/api/classes/AgentChain

Details the methods available on an AgentChain instance, specifically the 'push' method for adding tools.

```APIDOC
## push(tool: ToolChain)

### Description
Adds a ToolChain to the agent's list of tools.

### Method
`push`

### Parameters
#### Path Parameters
- **tool** (ToolChain) - Required - The ToolChain to add to the chain.
```

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

### Agent Constructors

Source: https://fellou.ai/eko/docs/api/classes/Agent

Information about the Agent class constructors.

```APIDOC
## Constructors

### constructor

- **Signature**: `new Agent(params: AgentParams): Agent`
- **Description**: Initializes a new instance of the Agent class.
- **Parameters**:
  - `params` (AgentParams) - Required - The parameters required for agent initialization.
```

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

### SimpleSseMcpClient Methods

Source: https://fellou.ai/eko/docs/api/classes/SimpleSseMcpClient

Provides methods for interacting with the EKO API via SSE. This includes connecting, checking connection status, listing available tools, executing tools, and handling incoming messages.

```typescript
callTool(param: McpCallToolParam): Promise<ToolResult>
```

```typescript
close(): Promise<void>
```

```typescript
connect(): Promise<void>
```

```typescript
isConnected(): boolean
```

```typescript
listTools(param: McpListToolParam): Promise<McpListToolResult>
```

```typescript
onmessage(data: SseEventData): void
```

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

### Instantiate BaseChatAgent

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Demonstrates how to create a new instance of the BaseChatAgent class. It accepts optional arguments for language models, external tools, and an MCP client.

```typescript
new BaseChatAgent(
  llms?: string[],
  ext_tools?: Tool[],
  mcpClient?: IMcpClient
): BaseChatAgent
```

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

### Create Web Content with HTML, CSS, JavaScript

Source: https://fellou.ai/eko/docs/getting-started/eko/installation

The AI Browser can generate code for web pages and applications, including websites, games, and interactive models, using HTML, CSS, and JavaScript.

```html
<html>
<head>
  <title>Sample Page</title>
  <link rel="stylesheet" href="style.css">
</head>
<body>
  <h1>Hello, World!</h1>
  <script src="script.js"></script>
</body>
</html>
```

```css
h1 {
  color: blue;
  text-align: center;
}
```

```javascript
console.log("JavaScript loaded successfully!");

document.querySelector('h1').addEventListener('click', () => {
  alert('Heading clicked!');
});
```

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

### BaseFileAgent Methods

Source: https://fellou.ai/eko/docs/api/classes/BaseFileAgent

Contains methods for running the agent, either with or without context.

```APIDOC
## Methods

### run

Executes the agent's logic with the provided context and agent chain.

#### Parameters

* `context` (Context): The context for the agent's operation.
* `agentChain` (AgentChain): The agent chain to be executed.

#### Returns

* `Promise<string>`: A promise that resolves with the result of the agent's execution.

### runWithContext

Executes the agent's logic with the provided agent context.

#### Parameters

* `agentContext` (AgentContext): The agent context containing all necessary information.
* `mcpClient` (IMcpClient, Optional): An instance of the MCP client. Defaults to undefined.
* `maxReactNum` (number, Optional): The maximum number of reactions to process. Defaults to 100.

#### Returns

* `Promise<string>`: A promise that resolves with the result of the agent's execution.
```

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

### Browser Agent - Scrolling and Utility Tools

Source: https://fellou.ai/eko/docs/agents/browser-agent

Tools for scrolling the page and managing execution flow.

```APIDOC
## POST /browser_agent/scroll_mouse_wheel

### Description
Scrolls the page vertically by a specified amount.

### Method
POST

### Endpoint
/browser_agent/scroll_mouse_wheel

### Parameters
#### Request Body
- **amount** (number) - Required - The amount to scroll vertically. Positive values scroll down, negative values scroll up.

### Request Example
```json
{
  "amount": 100
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the scroll operation.

#### Response Example
```json
{
  "status": "scrolled successfully"
}
```

## POST /browser_agent/wait

### Description
Pauses the execution for a specified duration. Useful for waiting for content to load or animations to complete.

### Method
POST

### Endpoint
/browser_agent/wait

### Parameters
#### Request Body
- **milliseconds** (number) - Required - The duration to wait in milliseconds.

### Request Example
```json
{
  "milliseconds": 2000
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates that the wait operation has completed.

#### Response Example
```json
{
  "status": "wait completed"
}
```
```

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

### Agent Constructor

Source: https://fellou.ai/eko/docs/api/classes/Agent

Initializes a new instance of the Agent class. It requires AgentParams as input.

```typescript
new Agent(params: AgentParams): Agent
```

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

### Browser Agent Scrolling and Utility Tools

Source: https://fellou.ai/eko/docs/agents/browser-agent

Provides tools for scrolling the mouse wheel and implementing delays. The `wait` tool is crucial for synchronizing automation with page load times or dynamic content.

```JavaScript
scroll_mouse_wheel(amount: number)
wait(milliseconds: number)
```

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

### Build Custom AI Agents with Eko 2.0 Framework (JavaScript)

Source: https://fellou.ai/eko/docs/getting-started/quickstart/installation

This snippet demonstrates the use of the open-source Eko 2.0 framework to build custom AI agents. It highlights the hybrid approach of using natural language and JavaScript for creating powerful, intervenable agents within the Fellou AI Browser.

```javascript
import { Agent } from '@eko/framework';

class CustomAgent extends Agent {
  async run(context) {
    // Use natural language and JavaScript to define agent behavior
    const response = await this.llm.invoke('Summarize the provided document.', { context });
    return response;
  }
}

export default CustomAgent;

```

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

### EKO API - initContext

Source: https://fellou.ai/eko/docs/api/classes/Eko

Initializes the context for a given workflow.

```APIDOC
## initContext(
  workflow: Workflow,
  contextParams?: Record<string, any>
): Promise<Context>

### Description
Initializes the context for a workflow.

### Parameters
* **workflow** (Workflow) - The workflow for which to initialize the context.
* **contextParams** (Record<string, any>, optional) - Optional parameters for context initialization.

### Returns
Promise<Context> - A promise that resolves with the initialized Context.
```

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

### Instantiate Custom Agent with Tools (JavaScript)

Source: https://fellou.ai/eko/docs/agents/custom-agent

Demonstrates the direct instantiation of a custom Agent in the Eko framework. This method involves creating a new Agent instance, providing a name, description, and a list of tools to define its functionality. It's a straightforward way to create agents for specific tasks.

```javascript
const agent = new Agent({
  name: "File Agent",
  description: "This agent is used to manage files.",
  tools: [
    // List of tools defined here
  ]
});
```

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

### Chain Constructor

Source: https://fellou.ai/eko/docs/api/classes/Chain

Initializes a new instance of the Chain class.

```APIDOC
## Constructors

### constructor

* `new Chain(taskPrompt: string): Chain`

#### Parameters

* `taskPrompt` (string) - The prompt for the task.

#### Returns

* `Chain` - A new instance of the Chain class.
```

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

### SimpleSseMcpClient.onmessage

Source: https://fellou.ai/eko/docs/api/classes/SimpleSseMcpClient

Handles incoming messages from the SSE server.

```APIDOC
## Methods onmessage

### Description
Handles incoming messages from the SSE server.

### Method
`POST`

### Endpoint
`/onmessage`

### Parameters
#### Request Body
- **data** (SseEventData) - Required - The data of the incoming SSE event.
```

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

### Browser Agent Navigation Tools

Source: https://fellou.ai/eko/docs/agents/browser-agent

Provides tools for navigating web pages, including going back, switching tabs, and retrieving all open tabs. These are essential for managing browser sessions during automated tasks.

```JavaScript
navigate_to(url: string)
go_back()
switch_tab(index: number)
get_all_tabs()
```

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

### Chain Methods

Source: https://fellou.ai/eko/docs/api/classes/Chain

Explains the methods available for interacting with the Chain class.

```APIDOC
## Methods

### addListener

* `addListener(callback: Callback): void`

#### Parameters

* `callback` (Callback) - The callback function to add.

#### Returns

* `void`

### push

* `push(agent: AgentChain): void`

#### Parameters

* `agent` (AgentChain) - The AgentChain to add to the chain.

#### Returns

* `void`

### removeListener

* `removeListener(callback: Callback): void`

#### Parameters

* `callback` (Callback) - The callback function to remove.

#### Returns

* `void`
```

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

### AgentContext Constructor

Source: https://fellou.ai/eko/docs/api/classes/AgentContext

Initializes a new instance of the AgentContext class with the provided context, agent, and agentChain.

```APIDOC
## new AgentContext

### Description
Initializes a new instance of the AgentContext class.

### Method
CONSTRUCTOR

### Endpoint
N/A

### Parameters
#### Path Parameters
N/A

#### Query Parameters
N/A

#### Request Body
N/A

### Request Example
```typescript
// Assuming Context, Agent, and AgentChain are defined elsewhere
const myContext: Context = { /* ... */ };
const myAgent: Agent = { /* ... */ };
const myAgentChain: AgentChain = { /* ... */ };

const agentContext = new AgentContext(myContext, myAgent, myAgentChain);
```

### Response
#### Success Response (200)
N/A (Constructor does not return a response in the traditional sense, it creates an object).

#### Response Example
N/A
```

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

### Class ForeachTaskTool

Source: https://fellou.ai/eko/docs/api/classes/ForeachTaskTool

Provides details on the ForeachTaskTool class, including its constructor, properties, and the execute method.

```APIDOC
## Class ForeachTaskTool

### Description
This class, ForeachTaskTool, is part of the EKO API and implements the Tool interface. It is designed for iterating over tasks.

### Methods
#### execute
This method executes the task with the provided arguments and agent context.

##### Parameters
- **args** (Record<string, unknown>) - Required - A record of unknown key-value pairs representing the arguments for the task.
- **agentContext** (AgentContext) - Required - The context of the agent performing the task.

##### Returns
Promise<ToolResult> - A promise that resolves to a ToolResult object.

### Properties
#### description
- **description** (string) - A description of the tool.
#### name
- **name** (string) - The name of the tool, typically set to TOOL_NAME.
#### parameters
- **parameters** (JSONSchema7) - The JSON schema defining the parameters for the tool.
```

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

### BaseTimerAgent Constructor

Source: https://fellou.ai/eko/docs/api/classes/BaseTimerAgent

Initializes a new instance of the BaseTimerAgent.

```APIDOC
## new BaseTimerAgent

### Description
Initializes a new instance of the BaseTimerAgent.

### Method
CONSTRUCTOR

### Parameters
#### Path Parameters
* None

#### Query Parameters
* None

#### Request Body
* None

### Request Example
```json
// No request body for constructor
```

### Response
#### Success Response (Instance)
* **BaseTimerAgent** - An instance of the BaseTimerAgent.

### Response Example
```json
// Constructor does not return JSON, it returns an instance
```
```

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

### VariableStorageTool execute Method

Source: https://fellou.ai/eko/docs/api/classes/VariableStorageTool

Executes the VariableStorageTool with provided arguments and agent context.

```APIDOC
## Methods
### execute
  * execute(
args: Record<string, unknown>,
agentContext: AgentContext,
): Promise<ToolResult>
#### Parameters
    * args: Record<string, unknown>
    * agentContext: AgentContext
#### Returns Promise<ToolResult>
```

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

### Agent Methods

Source: https://fellou.ai/eko/docs/api/classes/Agent

Details on the methods available for the Agent class.

```APIDOC
## Methods

### run

- **Signature**: `run(context: Context, agentChain: AgentChain): Promise<string>`
- **Description**: Executes the agent's primary function with the provided context and agent chain.
- **Parameters**:
  - `context` (Context) - Required - The execution context for the agent.
  - `agentChain` (AgentChain) - Required - The agent chain to be executed.
- **Returns**: `Promise<string>` - A promise that resolves with the string output of the agent's execution.

### runWithContext

- **Signature**: `runWithContext(agentContext: AgentContext, mcpClient?: IMcpClient, maxReactNum?: number): Promise<string>`
- **Description**: Executes the agent's function with detailed context, optionally including an MCP client and a maximum reaction number.
- **Parameters**:
  - `agentContext` (AgentContext) - Required - The detailed context for the agent's operation.
  - `mcpClient` (IMcpClient) - Optional - The MCP client to be used.
  - `maxReactNum` (number) - Optional, Defaults to 100 - The maximum number of reactions allowed during execution.
- **Returns**: `Promise<string>` - A promise that resolves with the string output of the agent's execution.
```

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

### BaseChatAgent Methods

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Documentation for the methods available in the BaseChatAgent class.

```APIDOC
## Method run

### Description
Executes the agent's run logic.

### Parameters
* **context** (Context) - The context for the agent.
* **agentChain** (AgentChain) - The agent chain to execute.

### Returns
* **Promise<string>** - A promise that resolves with the result string.
```

```APIDOC
## Method runWithContext

### Description
Executes the agent's logic with provided context and optional parameters.

### Parameters
* **agentContext** (AgentContext) - The agent context.
* **mcpClient** (IMcpClient) - Optional - MCP client instance.
* **maxReactNum** (number) - Optional - Maximum number of reactions. Defaults to 100.

### Returns
* **Promise<string>** - A promise that resolves with the result string.
```

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

### Browser Agent - Element Interaction Tools

Source: https://fellou.ai/eko/docs/agents/browser-agent

Tools for interacting with elements on a web page using their indexes.

```APIDOC
## POST /browser_agent/input_text

### Description
Inputs text into an element identified by its index. Optionally presses Enter after typing.

### Method
POST

### Endpoint
/browser_agent/input_text

### Parameters
#### Request Body
- **index** (number) - Required - The index of the element to input text into.
- **text** (string) - Required - The text to input.
- **enter** (boolean) - Optional - Whether to press Enter after typing.

### Request Example
```json
{
  "index": 5,
  "text": "Hello, world!",
  "enter": true
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the text input operation.

#### Response Example
```json
{
  "status": "text input successful"
}
```

## POST /browser_agent/click_element

### Description
Performs a click action on an element identified by its index. Supports specifying the number of clicks and the mouse button.

### Method
POST

### Endpoint
/browser_agent/click_element

### Parameters
#### Request Body
- **index** (number) - Required - The index of the element to click.
- **num_clicks** (number) - Optional - The number of times to click the element.
- **button** (string) - Optional - The mouse button to use ('left', 'right', or 'middle'). Defaults to 'left'.

### Request Example
```json
{
  "index": 10,
  "num_clicks": 2,
  "button": "right"
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the click operation.

#### Response Example
```json
{
  "status": "element clicked successfully"
}
```

## POST /browser_agent/hover_to_element

### Description
Moves the mouse pointer over an element identified by its index.

### Method
POST

### Endpoint
/browser_agent/hover_to_element

### Parameters
#### Request Body
- **index** (number) - Required - The index of the element to hover over.

### Request Example
```json
{
  "index": 15
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the hover operation.

#### Response Example
```json
{
  "status": "hovered to element successfully"
}
```

## POST /browser_agent/scroll_to_element

### Description
Scrolls the page to bring the specified element into view.

### Method
POST

### Endpoint
/browser_agent/scroll_to_element

### Parameters
#### Request Body
- **index** (number) - Required - The index of the element to scroll to.

### Request Example
```json
{
  "index": 20
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the scroll operation.

#### Response Example
```json
{
  "status": "scrolled to element successfully"
}
```

## POST /browser_agent/select_option

### Description
Selects an option from a dropdown or select element by its index.

### Method
POST

### Endpoint
/browser_agent/select_option

### Parameters
#### Request Body
- **index** (number) - Required - The index of the select element.
- **option_value** (string) - Required - The value of the option to select.

### Request Example
```json
{
  "index": 25,
  "option_value": "option2"
}
```

### Response
#### Success Response (200)
- **status** (string) - Indicates the success of the select option operation.

#### Response Example
```json
{
  "status": "option selected successfully"
}
```

## GET /browser_agent/get_select_options

### Description
Retrieves the available options for a select element identified by its index.

### Method
GET

### Endpoint
/browser_agent/get_select_options

### Parameters
#### Query Parameters
- **index** (number) - Required - The index of the select element.

### Response
#### Success Response (200)
- **options** (array) - A list of available options for the select element.

#### Response Example
```json
{
  "options": [
    { "value": "option1", "label": "Option 1" },
    { "value": "option2", "label": "Option 2" }
  ]
}
```
```

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

### Build Custom AI Agents with Natural Language and JavaScript

Source: https://fellou.ai/eko/docs/getting-started/eko/installation

Developers can use the open-source Eko 2.0 framework to build custom AI agents. These agents can run within the Fellou AI Browser by combining natural language instructions with JavaScript code.

```javascript
// Example of a hybrid agent using natural language intent and JavaScript logic

function processUserInput(input) {
  if (input.toLowerCase().includes("fetch data from api")) {
    fetchData();
  } else if (input.toLowerCase().includes("calculate sum")) {
    // Assume input is like "calculate sum of 5 and 10"
    const numbers = input.match(/\d+/g).map(Number);
    if (numbers.length === 2) {
      console.log(`Sum: ${numbers[0] + numbers[1]}`);
    }
  } else {
    console.log(`Processing: ${input}`);
  }
}

async function fetchData() {
  try {
    const response = await fetch('https://api.example.com/data');
    const data = await response.json();
    console.log('Data fetched:', data);
  } catch (error) {
    console.error('Error fetching data:', error);
  }
}

// Simulate user input
processUserInput("Fetch data from API");
processUserInput("Calculate sum of 25 and 75");
processUserInput("Log this message");

```

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

### BaseChatAgent Constructors

Source: https://fellou.ai/eko/docs/api/classes/BaseChatAgent

Provides details about the constructors available for the BaseChatAgent class.

```APIDOC
## Constructor BaseChatAgent

### Description
Initializes a new instance of the BaseChatAgent class.

### Parameters
* **llms** (string[]) - Optional - Array of language models.
* **ext_tools** (Tool[]) - Optional - Array of external tools.
* **mcpClient** (IMcpClient) - Optional - MCP client instance.

### Returns
* **BaseChatAgent** - An instance of the BaseChatAgent class.
```

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

### AgentChain Constructor

Source: https://fellou.ai/eko/docs/api/classes/AgentChain

Initializes a new instance of the AgentChain class with a specified WorkflowAgent.

```APIDOC
## new AgentChain(agent: WorkflowAgent)

### Description
Initializes a new instance of the AgentChain class.

### Method
Constructor

### Parameters
#### Path Parameters
- **agent** (WorkflowAgent) - Required - The WorkflowAgent to associate with this chain.
```

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

### EKO API - generate

Source: https://fellou.ai/eko/docs/api/classes/Eko

Generates a workflow based on a task prompt.

```APIDOC
## generate(
  taskPrompt: string,
  taskId?: string,
  contextParams?: Record<string, any>
): Promise<Workflow>

### Description
Generates a workflow based on a task prompt.

### Parameters
* **taskPrompt** (string) - The prompt describing the task to generate a workflow for.
* **taskId** (string, optional) - An optional task ID to associate with the generation.
* **contextParams** (Record<string, any>, optional) - Optional parameters for context.

### Returns
Promise<Workflow> - A promise that resolves with the generated Workflow.
```
Tokens:
Raw