> ## Documentation Index
> Fetch the complete documentation index at: https://docs.themcp.company/llms.txt
> Use this file to discover all available pages before exploring further.

# Quick start

> Add observability to your app in two lines of code

## TypeScript

```typescript theme={null}
import { init } from "mcp-tap";

init({ apiKey: "tap_your_project_key" });

// That's it. All supported SDK calls are now captured automatically.
```

## Python

```python theme={null}
import mcp_tap

mcp_tap.init(api_key="tap_your_project_key")

# All supported SDK calls are now captured automatically.
```

After calling `init()`, every call to Anthropic, OpenAI, MCP `callTool`, Composio `executeAction`, and HTTP requests to known AI APIs are intercepted, measured, and sent to the mcp-tap dashboard.

## What happens next

Once initialized, mcp-tap captures events based on which [integration path](/guides/sdk-patching) is active:

| Path                                                     | How it works                                           | Best for                                                 |
| -------------------------------------------------------- | ------------------------------------------------------ | -------------------------------------------------------- |
| [SDK patching](/guides/sdk-patching)                     | `init()` monkey-patches installed AI SDKs              | Application code that calls LLM/tool APIs directly       |
| [Server instrumentation](/guides/server-instrumentation) | `instrumentServer()` wraps MCP Server request handlers | MCP server authors who want to log incoming tool calls   |
| [Stdio proxy](/guides/stdio-proxy)                       | `mcp-tap` binary sits between an MCP client and server | IDE integrations, zero-code monitoring of any MCP server |