Push code. Get a governed endpoint.

Managed hosting for MCP servers with the operational basics handled end-to-end. Live in 60 seconds.

60 second deploys
Branch previews
Scale to zero

Where the “real world” details show up

A surprising amount of MCP work is actually long-running: background tasks, multi-step workflows, subagent delegation, and other operations that shouldn't block an agent in a chat loop.

Horizon Deploy supports the MCP background tasks protocol so agents can kick off work asynchronously and keep moving.

How It Works

Connect your repo, push, deploy

Point Horizon Deploy at your GitHub repository. Every push to main automatically builds, containerizes, and deploys your MCP server. Preview deployments spin up for pull requests. No YAML configs, no deployment scripts, no infrastructure to manage.

Get started
# Write your MCP server
$ cat server.py
from fastmcp import FastMCP
mcp = FastMCP("my-tools")
@mcp.tool()
def query_database(sql: str) -> dict:
return db.execute(sql)
# Push to GitHub
$ git push origin main
# Your server is live in 60 seconds
# https://my-tools.horizon.prefect.io

What You Get

Operational basics, handled end-to-end

CI/CD integration

Push to main, server updates. Open a PR, get a preview deployment. Versioning and rollbacks built in.

OAuth 2.1 included

Dynamic client registration out of the box. Required for Claude.ai and enterprise AI applications.

Scaling

Handles traffic spikes automatically. Scales to zero when unused—pay only for what you use.

Monitoring

See who's requesting what, which requests are timing out, and how your servers are performing.

Background Tasks Protocol

Long-running work without blocking the chat loop

Thanks to FastMCP, Horizon supports the MCP background tasks protocol. Agents can kick off multi-step workflows, subagent delegation, and other long-running operations asynchronously.

The agent keeps moving while the work happens in the background.

Deploy your first MCP server

Free hosting for personal servers. No credit card required.