# Service API Overview

Source: https://langbot.app/docs/en/tags/readme

LangBot provides a set of RESTful APIs for external services to integrate and automate management. These APIs are authenticated with API keys, which are a subset of the HTTP API.

> warning:
>
> Service API is currently in Beta and some APIs may change frequently.

## Get API Key

Click the `API Keys` button in the sidebar of the WebUI to create an API Key.

![create\_api\_key](https://langbot.app/docs/images/zh/api/create_api_key.png)

After creation, copy the API Key and save it.

![copy\_api\_key](https://langbot.app/docs/images/zh/api/copy_api_key.png)

## Authentication

All Service APIs support two authentication methods:

- **API Key**: Via `X-API-Key: <your-key>` or `Authorization: Bearer <your-key>` header
- **User Token**: Via `Authorization: Bearer <user-token>` header (obtained after login)

## Use API

The left sidebar lists all available APIs. You can fill in your instance address (the port address exposed by the backend, default is `http://localhost:5300`) and API Key, then click the `Try it out` button to call the API.

## Core Concepts

### Provider + Model Architecture

LangBot uses a **Provider + Model** two-layer architecture for AI model management:

- **Provider**: Defines the API endpoint (`base_url`), requester type, and API keys. A single provider can be associated with multiple models.
- **Model**: Associated with a provider (via `provider_uuid`), defining the specific model name and parameters. Models inherit the provider's API endpoint and keys.

```
Provider (OpenAI)
├── base_url: https://api.openai.com/v1
├── api_keys: [sk-xxx]
├── LLM: gpt-4o
├── LLM: gpt-4o-mini
└── Embedding: text-embedding-3-small
```

This architecture allows you to manage multiple models under a single provider, sharing API configuration and avoiding redundant key entries.

### Requester

A requester defines the communication protocol with AI services. Use `GET /api/v1/provider/requesters` to list available requesters. Common requesters include:

- `openai-chat-completions` — OpenAI-compatible chat completions API (also works with SiliconFlow, DeepSeek, etc.)
- `anthropic-messages` — Anthropic Claude API
- `openai-embeddings` — OpenAI-compatible embeddings API

### Async Tasks

Some operations (plugin install/delete/upgrade) run asynchronously and return a `task_id`. You can track progress in the WebUI.