Glossary · Deploying and hosting
Self-hosting
Self-hosting means running software on infrastructure whose administration you manage or control. In AI, this may involve a model, interface, agent, or MCP server, but it does not necessarily mean that every component runs on your own computer or that all information stays within a single local environment.
Self-hosting: the explanation
You install the necessary components on your computer, an organizational server, or rented infrastructure. You manage configuration and, depending on the arrangement, updates, access, and availability. Self-hosting therefore describes an operational responsibility as well as a deployment choice. An application you host can still call a remote model. To understand the setup, identify the interface, model, storage, and connected services separately. The location of one component does not establish where the entire system performs its processing or keeps its data.
This approach can help you customize a tool, control some data flows, or integrate an application into your existing environment. It may also support local operation when all required components permit it. In return, you need suitable computing resources and operational skills. Requirements vary greatly between a lightweight interface and a model running on your own hardware. Check the licenses too: being able to download a component does not mean that every use is unrestricted or suitable for your organizational requirements.
More control does not automatically produce better security, lower cost, or complete confidentiality. Dependencies, logs, backups, and outbound connections still matter. For example, a self-hosted MCP server may pass information to an external service through the tools it exposes. Examine the complete data path and maintenance responsibilities before choosing a setup. Also consider how you will apply security fixes, recover from failures, and restrict agent permissions. Hosting decisions are useful only when they match the way the whole system operates.
A concrete example
Your community organization wants to search an internal document collection. You install an interface and retrieval system on a server you administer. Before describing the setup as entirely local, you check where the model runs and which document excerpts it receives. You define access rules, backups, and an update process. You then test with nonsensitive material to confirm that the actual information flows match the arrangement you intended to deploy.
Why it matters
Self-hosting matters when you want greater control over operations, integrations, or selected data processing. It also creates responsibilities and costs beyond initial installation. Compare resource needs, maintenance effort, licenses, and external connections. The right setup should fit your skills and constraints. Where an application is installed is only one consideration; it does not by itself guarantee confidentiality, security, or complete independence from other services used by the agent.
Common misconceptions
- Hosting an interface yourself does not mean that the model runs locally, because the interface may send requests to a remote API and rely on external services for search, storage, or other functions available to the agent.
- Self-hosting is not automatically free or safer, because you become responsible for resources, updates, access controls, and backups, while a poor configuration can introduce risks even when you have more control over the software and infrastructure.
Related terms: deploying and hosting
- local AI
- Local AI is artificial intelligence that runs its calculations on your device or on infrastructure controlled by your organisation.
- LLM observability
- LLM observability is the set of methods used to understand an application built with language models through its recorded events and measurements.
- MLOps
- MLOps is a set of practices for developing, deploying, and maintaining machine learning systems in an organized way.
- OpenAI-compatible API
- An OpenAI-compatible API is a programming interface that reproduces some or all of an OpenAI API's exchange formats so existing software can connect to it.
Where this site covers it
The term appears on 581 pages of this site. A few of them:
- n8n and AI: A complete guide to nodes, agents, memory, hosting, and costs Guide
- Choosing an MCP server: trust, maintenance, permissions, and local or remote hosting Guide
- OpenClaw reference guide: installation, configuration, models, skills, messaging, security, and costs Guide
- Getting started with an AI agent: choosing tools, setup, first tasks, and safeguards Guide
- Connect an AI agent to Telegram, WhatsApp, Slack, or Discord: methods and risks Guide
- Deploying AI agents in business: use cases, selection, pilots, governance, and measurement Guide
- Deep research with AI: tools, methods, and source verification Guide
- LangGraph reference guide: state graphs, memory, tools, and deployment Guide
Go further
Frequently asked questions
Do you need to own a server?
No. You can administer a local computer or rented infrastructure. What matters is understanding which responsibilities you control and which remain with a provider. Your choice depends on the components, resource requirements, and constraints of the intended setup.
Can it work without internet access?
Yes, if every necessary component works locally and the configuration does not depend on remote services. Check this for the model, tools, any licensing requirements, and data sources. Self-hosting alone does not establish that offline operation is possible.
Will your data stay with you?
Not automatically. Review model requests, connected tools, logs, and backups. A self-hosted application can still transmit information elsewhere. You need to inspect the complete configuration and its connections to understand where your data is processed and stored.
Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score