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

# Context Hub ChatGPT Plugin

> A skills-only ChatGPT plugin that sets up Pipecat Context Hub and runs Setup, Explore, Build, and Deploy workflows for Pipecat agents.

The [Pipecat Context Hub plugin](https://github.com/pipecat-ai/pipecat-context-hub/tree/main/plugins/pipecat-context-hub) is a skills-only ChatGPT plugin for exploring, building, and deploying voice and multimodal Pipecat agents. It provides Setup, Explore, Build, and Deploy workflows grounded in local [Context Hub](/api-reference/context-hub/overview) CLI queries.

## Install the Plugin

[Install Pipecat Context Hub in ChatGPT](https://chatgpt.com/plugins/plugins_6ac6b44c8d1881919f9dc3ffefbc6007).

The plugin is a wrapper around the Pipecat CLI and can be invoked from the ChatGPT desktop application. Installing the plugin adds its skills and branding assets. When onboarding is requested, Setup can install or reuse the Pipecat CLI, the Context Hub runtime, the Cloud extension, and a local index in the selected execution environment. It can also start Pipecat Cloud account and login setup; you approve the changes and complete browser authentication yourself. It does not install credentials or provision Cloud infrastructure. Some current Pipecat CLI distributions bundle Context Hub; otherwise Setup installs it alongside the CLI. Build and Deploy still require local command execution and, for Cloud workflows, an authenticated Pipecat Cloud environment. The plugin uses Context Hub CLI queries and does not require MCP.

After activating the plugin, the chat might ask **“Set up Pipecat Context Hub”**. The plugin
checks the local CLI and index, then uses grounded Context Hub queries for
Explore, Build, and Deploy.

For other coding clients, [set up Context Hub directly](/api-reference/context-hub/overview#setup). That setup does not install or configure the ChatGPT plugin.

See the plugin [README](https://github.com/pipecat-ai/pipecat-context-hub/blob/main/plugins/pipecat-context-hub/README.md) and [changelog](https://github.com/pipecat-ai/pipecat-context-hub/blob/main/plugins/pipecat-context-hub/CHANGELOG.md) for packaging and release details.

## Inline Recommendation

All three surfaces can be explicitly installed, but only the ChatGPT plugin
can be surfaced inline by ChatGPT or Codex on the desktop app. The plugin may
be recommended while a user is discussing how to build an agent; CLI and MCP
are not inline recommendation surfaces.

| Surface | Explicit installation | Inline recommendation in the ChatGPT/Codex desktop app |
| - | - | - |
| [CLI](/api-reference/context-hub/overview#cli-lookups) | Install `pipecat-ai[cli]` manually | No |
| [MCP](/api-reference/context-hub/overview#mcp-server) | Run `pipecat context-hub install` or register the server manually | No |
| ChatGPT plugin | [Install Pipecat Context Hub in ChatGPT](https://chatgpt.com/plugins/plugins_6ac6b44c8d1881919f9dc3ffefbc6007) | Yes, host-controlled |

<Note>
  Inline plugin surfacing, activation, and approval behavior remain controlled
  by the host and the user. Explicit installation from the link above is
  separate from inline recommendation, and a recommendation is not guaranteed
  for every conversation. The plugin cannot change ChatGPT approval settings.
</Note>

## How the Plugin Works

The plugin's normal workflow is:

1. **Setup** checks whether the host can run the required commands, whether
   Context Hub is installed, and whether a usable local index exists.
2. **Explore** retrieves relevant documentation, examples, and API details before
   making Pipecat-specific recommendations.
3. **Build** discovers the installed Pipecat CLI, validates the requested
   scaffold options, and creates or adapts an application after an explicit
   build request.
4. **Deploy** checks the application and Cloud target, prepares an exact
   deployment and secret-change proposal, and waits for approval before making
   deployment or secret mutations.

The agent should retrieve the relevant source before making framework-specific
API claims. Retrieval is evidence for implementation; it is not a substitute
for running tests or validating provider credentials.

The plugin reuses healthy installations and indexes. It does not silently
refresh an index, replace an installed CLI, upload credentials, or deploy an
application.

For example, a request to build a Daily bot with TTS and STT should lead the
agent to find the relevant concepts, example, API surface, and lifecycle status
before writing code with the [Context Hub tools](/api-reference/context-hub/overview#tools):

```text theme={null}
search_docs("Daily + TTS + STT")
search_examples("Daily TTS STT", domain="backend")
search_api("DailyTransport")
check_deprecation("PipelineTask")
```

The plugin works from a standard Context Hub index, so
[custom sources](/api-reference/context-hub/overview#custom-sources) and
[index freshness checks](/api-reference/context-hub/overview#keeping-the-index-current)
apply to it as well.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.