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

# Choose a Hosting Model for MCP Servers

> Compare self-hosted, Pylee-managed, package-based, and hybrid deployment patterns.

# Hosting Models & Deployment Patterns

Use this guide to choose how your MCP server runs in development and production. Each option includes pros/cons and a minimal config example.

## Quick Comparison

| Option        | Setup | Control |        Cost |      Scale | Best For                 |
| ------------- | ----: | ------: | ----------: | ---------: | ------------------------ |
| Self-Hosted   |  High |    Full |     You pay | You manage | Large-scale products     |
| Pylee-Managed |   Low |  Shared | Usage-based |  Automatic | Most teams (coming soon) |
| Package-Based |   Low |    User |        Free |        N/A | OSS / local tools        |
| Hybrid        |   Med |   Mixed |      Varies |   Flexible | Enterprise solutions     |

<Tip>
  Prototype locally with **Package-Based**. For team usage, move to **Pylee-Managed** (when available) or **Self-Hosted**. Use **Hybrid** when different environments need different models.
</Tip>

***

## Self-Hosted (Remote HTTP/SSE)

You operate the server (cloud/infra of your choice), exposing an HTTP/SSE endpoint.

**When to choose**

* You need full control, custom scaling, or specific compliance controls.
* You already have an ops platform and observability in place.

**Pros**

* Maximum control and customization
* Fine-grained scaling and performance tuning

**Cons**

* DevOps burden (uptime, security, on-call)
* You pay infra costs directly

**Minimal listing example**

```json theme={null}
{
  "type": "remote",
  "url": "https://mcp.yourcompany.com",
  "transport": "sse"
}
```

***

## Pylee-Managed (Hosted) — *Early Access*

Pylee runs the server for you. You provide the package/endpoint; Pylee handles infra, scaling, and monitoring.

**When to choose**

* You want zero DevOps and fast team-wide adoption.
* You value integrated secrets, monitoring, and CDN.

**Pros**

* No infrastructure to manage
* Automatic scaling and built-in observability
* Integrated secrets management

**Cons**

* Less low-level control than self-hosted
* Availability phased (early access)

<Info>
  Interested in early access? Contact <a href="mailto:support@pyleeai.com">[support@pyleeai.com](mailto:support@pyleeai.com)</a>.
</Info>

***

## Package-Based (Local stdio)

Users run the server on their own machines (npm/PyPI/Docker). Great for development, personal workflows, and OSS.

**When to choose**

* Local development and prototyping
* Data must stay on the user’s machine

**Pros**

* No hosting cost
* Works offline; user data stays local

**Cons**

* Version fragmentation across machines
* Limited to local resources; not shared by default

**Minimal listing example**

```json theme={null}
{
  "type": "stdio",
  "package": "@yourorg/mcp-server",
  "command": "npx",
  "args": ["@yourorg/mcp-server"]
}
```

***

## Hybrid Patterns

Mix models by environment:

* **Dev**: Package-Based local stdio
* **Staging**: Pylee-Managed (or Self-Hosted small cluster)
* **Prod**: Self-Hosted with autoscaling and hard SLOs

**When to choose**

* Different teams/environments need different guarantees
* You’re migrating from local to managed/remote over time

**Pros**

* Right-sized tradeoffs by environment
* Smooth migration path

**Cons**

* More configurations to track
* Requires clear version and rollout discipline

***

## Decision Guide

* Need strict control/compliance/SLOs? → **Self-Hosted**
* Want zero DevOps and fastest org rollout while balancing security? → **Pylee-Managed** (when available)
* Prototyping or OSS tooling? → **Package-Based**
* Different needs across envs? → **Hybrid**

<Note>
  Whatever you choose, keep secrets out of code. Prefer environment variables or the Pylee Secrets Vault.
</Note>