Emerging category
AI agent hosting
The infrastructure that keeps autonomous AI agents running: always-on compute, persistent memory, and safe access to tools and the open web. Not a website that waits for visitors, but a worker that keeps going on its own.
What makes it different
A normal site serves a page when someone asks, then goes idle. An AI agent is a long-running process that thinks, calls tools and takes actions continuously. That means it needs somewhere that stays alive, remembers what it has done, and lets it reach out to APIs and the web, none of which a standard hosting plan is built for.
Do you even need a GPU?
Usually not. Most agents call a hosted model through an API, so the heavy lifting happens on the model provider's side and a modest CPU box is plenty. You only need a GPU if you are running the model itself locally rather than calling one over the network.
What an agent actually needs from a host
The things that separate somewhere an agent can really live from a plain web server.
Always-on compute
Agents are long-running processes, not request-and-response pages. The host has to keep them alive between actions, for hours or days.
Persistent state and memory
An agent needs to remember what it did and what it learned. Look for durable storage and state that survives restarts.
Open outbound access
Agents call APIs, tools and the live web. Tight egress rules or blocked ports will quietly break them.
Sandboxing and isolation
Agents run code and take actions you did not write by hand. Each one should be boxed in so a mistake cannot reach anything else.
Secrets management
API keys and tokens need to live somewhere safer than a plain environment variable in a shared box.
Scheduling and triggers
Cron, webhooks and queues let an agent wake on a timer or an event instead of running in a loop forever.
Observability and replay
When an agent does something odd you need the logs and the trace to see every step it took and why.
GPU access (optional)
Most agents call a model over an API and need no GPU. You only need one if you run the model locally.
Platforms for hosting AI agents
Places built to run autonomous agents, from sandboxes to full agent runtimes.
We are curating this list right now.
We only list platforms once we have verified what they actually offer. Run an agent platform, or know one worth listing?
Suggest a platform ›What people host agents to do
Common workloads that need an always-on home.
Research and monitoring
Agents that watch sources, gather information and report back on a schedule.
Coding agents
Long-running dev environments where an agent edits, runs and tests code in a sandbox.
Web automation
Headless-browser agents that navigate, fill forms and pull data from live sites.
Support and operations
Agents that triage tickets, answer questions and take routine actions for a team.
Multi-agent systems
Several agents running at once, talking to each other and dividing up the work.
AI agent hosting questions
It is the infrastructure that keeps an autonomous AI agent running: always-on compute, persistent memory, and safe access to tools and the open web. Unlike a normal website, which only does something when a visitor asks, an agent keeps working on its own, so it needs somewhere that stays alive and lets it act.
Normal hosting serves pages in response to requests and then goes idle. An agent is a long-running process that thinks, calls tools and acts continuously, so it needs persistent state, freedom to make outbound calls, sandboxing for the code it runs, and observability so you can see what it did. Those needs rarely show up on a standard shared-hosting feature list.
Yes, and many people do. A VPS gives you the always-on process, root access and outbound network an agent needs. The extras a dedicated agent platform adds are sandboxing, state, scheduling and replay out of the box, which you would otherwise build yourself on a VPS.
Usually not. Most agents call a hosted model through an API, so the heavy compute happens on the model provider's side. You only need a GPU if you are running the model itself locally rather than calling one over the network.
Always-on compute, persistent storage for memory and state, unrestricted outbound access, strong sandboxing, proper secrets handling, scheduling, and good logs so you can replay what the agent did. The checklist on this page walks through each one.
Running an agent on your own box?
Plenty of agents live happily on a VPS or cloud server: always-on, root access, open network. Compare real plans by power and price.