> For the complete documentation index, see [llms.txt](https://neirth.gitbook.io/openlobster/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://neirth.gitbook.io/openlobster/plugins.md).

# Plugins

Plugins extend OpenLobster with new AI providers, messaging channels, memory backends, and more.

Plugins are binary extensions that add new capabilities to OpenLobster. Each plugin runs as a separate subprocess and communicates with the core through a bidirectional JSON-RPC 2.0 protocol over standard input/output streams. Plugins can provide AI model access, messaging integrations, memory storage, audio synthesis, and secret management backends.

## Plugin types

OpenLobster supports five categories of plugins, each identified by a type field:

| Type        | Description                                                     | Typical exports        |
| ----------- | --------------------------------------------------------------- | ---------------------- |
| `ai`        | AI model providers (Anthropic, OpenAI, Ollama)                  | `chat`, `embed`        |
| `messaging` | User chat channels (Telegram, Discord, Slack, WhatsApp, Twilio) | `send`, `start`        |
| `memory`    | Knowledge graph backends (GML, Neo4j)                           | `get`, `set`, `query`  |
| `audio`     | Text-to-speech services (ElevenLabs)                            | `synthesize`           |
| `secrets`   | Secret storage backends (JSON file, OpenBao)                    | `get`, `set`, `delete` |

## How plugins work

Plugins communicate with the OpenLobster host through a bidirectional channel:

1. **Initialization**: On startup, the plugin receives a metadata request and responds with its identifying information.
2. **Configuration**: The host sends configuration parameters that the plugin stores for subsequent operations.
3. **Execution**: The plugin receives function calls with input data and returns results asynchronously.
4. **Callbacks**: The plugin can call the host at any time to send messages, request secrets, or query data.

This design keeps the plugin isolated from the core, so crashes or bugs in a plugin do not affect the main application. Plugins are always asynchronous and can initiate requests to the host at any time.

## Using plugins

Most users install pre-built plugins from the OpenLobster plugin registry. To add a new plugin:

1. Build or download the plugin binary for your platform.
2. Place it in the configured plugins directory.
3. Configure the plugin through the Settings panel.
4. Enable the plugin in the desired channels.

## Creating custom plugins

If you need a provider or integration not available in the registry, you can create your own plugin. Plugins are built using the Rust SDK, which handles the protocol implementation so you can focus on your plugin's specific functionality.

{% hint style="info" %}
The SDK provides the JSON-RPC communication layer, schema validation, hot-configuration management, and the ability to call back to the host, so you only need to implement your plugin's business logic.
{% endhint %}

## Pages in this section

* [Plugin Protocol](/openlobster/plugins/protocol.md) — Technical details of the STDIO/JSON-RPC transport
* [AI Providers](/openlobster/plugins/ai.md) — Specification for LLM and provider plugins
* [Messaging Gateways](/openlobster/plugins/messaging.md) — Specification for chat platform integrations
* [Memory Backends](/openlobster/plugins/memory.md) — Specification for knowledge and storage plugins
* [Audio & TTS](/openlobster/plugins/audio.md) — Specification for audio processing plugins
* [Secrets & Vaults](/openlobster/plugins/secrets.md) — Specification for sensitive data management
* [Creating Custom Plugins](/openlobster/plugins/create.md) — Build your own plugin from scratch


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://neirth.gitbook.io/openlobster/plugins.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.