zerolith.ioChecking…
Checking…
Function platform · AI-drivable · Hosted in France

From monolith
to zerolith.

A function platform built for AI agents, and just as simple to drive by hand.

For AI agents

Plug Claude, or any MCP agent, into your account. It writes, deploys and drives your functions without a key ever changing hands.

Connect an agent »
For developers

Write your function, deploy it from the UI or the API, call it from a plain URL.

Start coding »
Free credit€5.00at signup, no card required.≈ 10 million free function calls
Agent (MCP)claude · mcp
# just ask in plain language
 Deploy this function and make it public.

 deploy_function → hello (python)
 update_function → visibility: public
 https://hello-fn-a1b2c3.api.zerolith.io
CLI / curlhandler.py · curl
# main.py — your handler
def handler(request):
    return {"hello": request.query.get("name", "world")}

# deploy & invoke
$ curl -H "Authorization: Bearer $KEY" \
     https://hello-fn-a1b2c3.api.zerolith.io?name=neo
Why zerolith

The same foundation, by hand or by agent

What matters day to day, whether your functions are deployed by hand or by an agent.

An AI agent can do it all, and so can you

Plug in an MCP agent and automate everything, or stay on the dashboard and the API. Both paths reach the same functions, and there's no key to share.

Usage-based billing, scale to zero

An idle function costs nothing, whether the call comes from a human or an agent. You pay for the time your code runs, plus a small per-call fee.

Hosted in France 🇫🇷

Your functions and your data stay on servers run in France, at OVH, whatever the driving mode.

Private or public functions

Keep a function behind an API key, open it to the public for a webhook, or fire it on a schedule with cron. Usage is billed the same way in all three cases.

Multi-language today

Python and Node.js today, more languages after that. The contract is the same everywhere: a request goes in, a response comes out.

Strict isolation

Each account runs in its own Kubernetes space, walled off from the rest. That matters all the more when an autonomous agent is operating on your behalf.

How it works

Code to production, by hand or via an agent

Two paths that end in the same place: a function live in seconds.

Via an AI agent
01

Connect your agent over MCP, securely.

02

Ask it to deploy the function.

03

Call it from the agent or by its URL.

Via classic code
01

Write a function in Python or Node.js.

02

Deploy from the UI or the API.

03

Call its URL with your API key.

Pricing · Pay as you go

Pricing

You pay for memory, CPU and requests, by the second. Sizes are resource presets; they don't change the unit rate.

Memory
€0.000004
per GB-second of reserved memory
CPU
€0.0000125
per reserved vCPU-second
Requests
€0.0000004
per invocation
Free signup credit
€5.00
free at signup

Full pricing »

Ready to deploy?

Create an account, take the free credit, and deploy your first function straight away.

Get started »