> For the complete documentation index, see [llms.txt](https://docs.maiagent.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.maiagent.ai/tech/en/platform-development/environments.md).

# Platform Environments (SaaS/Private Cloud/On-Premises)

Platform environment comparison：differences in privacy, cost, and scalability across SaaS, private cloud, and on-premises deployments, and feature differences between private cloud and on-premises dep

MaiAgent offers multiple deployment environments to meet different customer needs. With SaaS on a public cloud, customers can get started quickly and benefit from maximum scalability with lower upfront costs. Customers who handle sensitive data can choose a dedicated private cloud deployment, which offers high privacy and a customizable environment. For greater security requirements, customers can choose an on-premises deployment. Although deployment is more complex and costly, it provides the highest privacy and a solution that accommodates regulatory restrictions.

|                        | **Public Cloud (SaaS)**                                                                                   | **Private Cloud**                                                                                                                                                                      | **On-Premises**                                                                                                                                                                                             |
| ---------------------- | --------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Platform location      | MaiAgent Cloud                                                                                            | Customer cloud (AWS, Azure, GCP, Oracle)                                                                                                                                               | Customer data center                                                                                                                                                                                        |
| Large language models  | <p><mark style="color:red;"><strong>All models</strong></mark></p><p>GPT series, Claude, Gemini, etc.</p> | <p><mark style="color:red;"><strong>Cloud models and self-hosted models</strong></mark></p><p>GPT series, Claude, Gemini, etc. (connected using the customer's own cloud API keys)</p> | <p>Open-weight models</p><p>gpt-oss, Gemma, Qwen, etc.<br><br>Can also connect to cloud models such as the GPT series, Claude, and Gemini (connections from the host to cloud services must be allowed)</p> |
| Response quality       | <mark style="color:red;">**Best**</mark>                                                                  | <mark style="color:red;">**Best**</mark>                                                                                                                                               | Good (can be best, depending on the model used)                                                                                                                                                             |
| Deployment complexity  | Standard                                                                                                  | Medium                                                                                                                                                                                 | High                                                                                                                                                                                                        |
| Upfront cost           | <mark style="color:red;">**Low**</mark>                                                                   | Medium                                                                                                                                                                                 | High                                                                                                                                                                                                        |
| Usage cost             | Usage-based                                                                                               | Usage-based                                                                                                                                                                            | <mark style="color:red;">**Fixed**</mark>                                                                                                                                                                   |
| Time to launch         | <mark style="color:red;">**Fast**</mark>                                                                  | Standard                                                                                                                                                                               | High                                                                                                                                                                                                        |
| Scalability            | <mark style="color:red;">**Highest**</mark>                                                               | High                                                                                                                                                                                   | Low                                                                                                                                                                                                         |
| Maintenance difficulty | Standard                                                                                                  | High                                                                                                                                                                                   | Highest                                                                                                                                                                                                     |
| Privacy                | Standard                                                                                                  | High                                                                                                                                                                                   | <mark style="color:red;">**Highest**</mark>                                                                                                                                                                 |
| Use cases              | General-purpose                                                                                           | Sensitive data, specific industries                                                                                                                                                    | Highly confidential data, regulatory restrictions                                                                                                                                                           |
| Pricing                | Fixed price                                                                                               | Fixed price                                                                                                                                                                            | Project-based quote                                                                                                                                                                                         |

## Feature Differences Between Private Cloud and On-Premises <a href="#private-cloud-vs-onprem-features" id="private-cloud-vs-onprem-features"></a>

Private cloud and on-premises environments use the same deployment package. The difference is the deployment category configured during installation:

* **Private cloud**: The host can connect to cloud services. Voice Agents, the call center, cloud speech recognition, and cloud connectors can be enabled as needed.
* **On-premises**: The environment is treated as unable to connect to cloud services, so all four features above are disabled. Speech-to-text uses offline speech recognition deployed on the host instead.

Compared with SaaS, private cloud and on-premises deployments share these differences:

* **Version**: Private cloud and on-premises deployments use quarterly releases (such as `2026.Q3`). New SaaS features are included in later quarterly releases. For example, Office online editing is currently available only in SaaS. For versions and upgrade procedures, see [Versions and Upgrades](/tech/en/platform-development/onprem-versioning-and-upgrades.md).
* **Models**: Customers host their own model services or connect to cloud models using their own API keys. See [Connecting Model Services for On-Premises Deployments](/tech/en/platform-development/onprem-model-services.md).

For the complete feature comparison and how to choose a deployment type (single-host Docker, distributed Docker, or Kubernetes), see [Deployment Types and Feature Comparison](/tech/en/platform-development/onprem-deployment-types.md).

## Can You Switch Environments Later? <a href="#switching-environments" id="switching-environments"></a>

Changing model compute (from cloud models to on-premises models, or vice versa) does not mean migrating the platform. You can change models at any time (except Embedding models; see [Connecting Model Services for On-Premises Deployments](/tech/en/platform-development/onprem-model-services.md#hybrid-cloud-setup)), while migrating the platform itself depends on the situation:

| From → To                                             | Support                                                                                                                                                                                                                                                                                                                                                                             |
| ----------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Private cloud ↔ On-premises (same deployment package) | The host's location does not matter. The only difference is the deployment category in the configuration, which is set during the first installation. To change it later, contact MaiAgent. See [Deployment Types and Feature Comparison](/tech/en/platform-development/onprem-deployment-types.md#private-cloud-vs-onprem)                                                         |
| Single-host/distributed Docker → Kubernetes           | The deployment package provides a migration procedure. First set up the Kubernetes cluster and data services, then migrate PostgreSQL, uploaded files, and search indexes (the cache is not migrated). Downtime is required during migration; schedule it with MaiAgent                                                                                                             |
| Change hosts (same deployment type)                   | Restore a built-in backup on the new host. See [Operations Essentials: Backup and Restore](/tech/en/platform-development/onprem-operations.md#backup-and-restore)                                                                                                                                                                                                                   |
| SaaS → Private cloud or on-premises                   | There is no one-click migration for an entire organization. Skills can be exported as skill packages and imported into the new environment. Knowledge base documents must be uploaded again, and Agent settings, conversation history, member accounts, and conversation platform integrations must be recreated. The scope and approach are assessed per project; contact MaiAgent |
| Private cloud or on-premises → SaaS                   | Same as above                                                                                                                                                                                                                                                                                                                                                                       |

Before migrating, confirm the feature differences in the target environment (for example, on-premises deployments do not have Voice Agents, a call center, or cloud connectors). See [Feature Differences Between Private Cloud and On-Premises](#private-cloud-vs-onprem-features).


---

# 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 by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.maiagent.ai/tech/en/platform-development/environments.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

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.