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

# How Parle Fits Together: Architecture Overview

> A high-level view of Parle: how your team reaches it, where it runs, and the AI providers, storage and business systems it connects to on your behalf.

Parle is a hosted service. Your team uses it from a browser, from Microsoft Teams or from the command line, and Parle connects outward to the AI models, file storage and business systems your organization chooses. There is nothing to install on your servers.

```mermaid theme={null}
flowchart LR
  subgraph people["Your team"]
    web["Web browser<br/>(installable app)"]
    teams["Microsoft Teams"]
    cli["Parle CLI<br/>(developers)"]
  end

  parle["<b>Parle</b><br/>Hosted on Microsoft Azure<br/>Includes Parle managed models"]

  subgraph models["AI models"]
    anthropic["Anthropic"]
    openai["OpenAI"]
    google["Google Gemini"]
    foundry["Your Azure AI Foundry"]
  end

  subgraph storage["File storage"]
    sp["SharePoint"]
    drive["Google Drive"]
  end

  subgraph identity["Sign-in"]
    entra["Microsoft Entra ID"]
    gws["Google Workspace"]
  end

  subgraph apps["Connected apps"]
    catalog["Confluence · Jira · Linear · Notion<br/>Miro · monday.com · Productive<br/>Fathom · Fireflies · Otter"]
    custom["Your own MCP servers"]
  end

  search["Web search (Brave)"]

  web --> parle
  teams --> parle
  cli --> parle
  parle --> models
  parle --> storage
  parle --> identity
  parle --> apps
  parle --> search
```

## How your team reaches Parle

> [card] Web app
>
> Parle runs in any modern browser and can be installed as an app on desktop and mobile. All of your work happens here.

> [card] Microsoft Teams
>
> Mention Parle in a Teams channel linked to a project, and the question becomes a conversation in that project, with the same budgets and rules as the web app.

> [card] Parle CLI
>
> A small command-line tool developers use to sign in and connect their AI coding tool to a project for usage tracking.

Every route arrives at the same service and is checked in the same way: who you are, which organization you belong to, and what your role allows in that project.

## Where Parle runs

Parle is a software-as-a-service product **hosted on Microsoft Azure**. We operate the whole platform, so you don't deploy, patch or scale anything.

* **Your organization's data is kept separate.** Every organization is isolated in the application from every other: conversations, files, members and usage are scoped to your organization.
* **Secrets stay server-side.** Provider API keys and integration credentials are stored in a dedicated secrets vault and used only by the Parle service when it makes a call. They are never sent to a browser.
* **Parle managed models run inside Parle's own Azure environment.** If your organization uses Parle's managed models rather than its own keys, your conversations go to models hosted within Parle's Azure environment, not to a separate AI company.

## What Parle connects to

Parle connects outward only to the services your organization has set up or switched on. For each one, see who it is for and whose credential is used in [Connected Services](https://docs.parlehub.com/architecture/connected-services.md).

| Category | Services | Whose credential |
| - | - | - |
| **AI models** | Anthropic, OpenAI, Google Gemini, your own Azure AI Foundry, or Parle managed models | Yours (bring your own key), or Parle's for managed models |
| **File storage** | Parle storage, SharePoint, Google Drive | Parle's for Parle storage; access you grant for SharePoint and Drive |
| **Sign-in** | Email and password, Microsoft Entra ID, Google Workspace | Parle's own sign-in, or your identity provider |
| **Connected apps** | Confluence, Jira, Linear, Notion, Miro, monday.com, Productive, Fathom, Fireflies, Otter, and any MCP server you register | The organization's, the project's or each user's own connection |
| **Web search** | Brave, only if you switch Web Search on | Parle's |

> [!NOTE]
>
> Parle does not use your content to train AI models, and it does not sell your data.

## Related

* [Connected Services](https://docs.parlehub.com/architecture/connected-services.md): the full list, with what each one receives
* [Security and Compliance Overview](https://docs.parlehub.com/security/overview.md)
* [Plans & Billing](https://docs.parlehub.com/plans-and-billing.md): which plans include each connection


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