CircuitNotion Cloudcompute for agents
CircuitNotion Cloud · The agent-friendly VPS

Servers your agents can run themselves.

The VPS cloud built for agentic companies. Spin up Linux machines and container apps from a dashboard, a REST API or an MCP server: your agent creates the box, deploys its code and hands you a public HTTPS URL, with no SSH keys or YAML involved.

Sign in with CircuitNotion Read the docs
REST + MCPFree plan includedPay with MoMo or card
mcp · circuitnotion-cloud
agent ▸ create_instance({ name: "support-bot", template: "python-3.12", size: "small" })
  ✓ running · 1 vCPU · 1 GB · sandboxed

agent ▸ write_file({ path: "/app/main.py", content: "from fastapi import …" })
agent ▸ exec_command({ command: "pip install fastapi uvicorn" })
  exit 0 · 6.4s

agent ▸ update_instance({ start_command: "uvicorn main:app --port $PORT" })
agent ▸ http_request({ method: "POST", path: "/v1/answer" })
  200 OK · 182 ms

live at https://support-bot-7k2qd.circuitnotion.com

Everything an agent needs to ship a service

MCP-native

Point Claude, or any MCP client, at one URL. 22 tools create machines, run commands, edit files and ship APIs.

Real Linux, full root

Ubuntu, Debian, Alpine, Python and Node with a persistent disk, a browser terminal, logs and live metrics.

HTTPS URL in seconds

Every instance is live at <name>.circuitnotion.com through Cloudflare, with TLS handled for you.

Isolated by default

Each instance runs in its own sandboxed container with dropped capabilities and hard CPU, memory and process limits.

Any Docker image or git repo

Run a public image as an app, or build a repository's Dockerfile on every deploy (paid plans).

Files, exec and logs over the API

Upload code, run commands, stream logs and call your app over HTTP, all with one API key.

Private networking

Your instances share a private network and reach each other by name. Nobody else's can.

Pay monthly with MoMo or card

Start on the free plan. Paid plans run 30 days per payment through CircuitNotion Pay, with no automatic charges.

Built for agentic companies

Why agents love it here

Most clouds were built for people clicking through consoles. This one is built for the agents your team ships: every response is shaped so a model can act on it, and every agent gets only the access it needs.

Docs agents can read

A one-call brief at /llms.txt and an OpenAPI 3.1 spec, so any model or framework can learn the API.

Errors with a next step

Every error has a stable code and a hint that tells the agent what to do, not an HTML page.

Safe to retry

Idempotency keys on create and ?wait=true on long operations, so agents never double-provision or poll blindly.

A key per agent

Scoped, expiring keys that can be locked to one instance, with every action logged under the key's name.

From sign-in to a live URL

  1. 1

    Sign in with CircuitNotion

    Use your CircuitNotion account. The free plan gives you one small instance straight away.

  2. 2

    Connect your agent

    Create an API key with the scopes it needs, then add the MCP server or call the REST API.

  3. 3

    Let it build and deploy

    The agent creates an instance, writes code, installs dependencies and starts it. You get an HTTPS URL.

Claude Code · add the MCP server
claude mcp add --transport http circuitnotion-cloud https://cloud.circuitnotion.com/mcp \
  --header "Authorization: Bearer cnc_..."

GPU instances for your agents' models

Dedicated NVIDIA GPUs, never time-shared, on the same API and MCP tools. Join the waitlist and tell us which GPUs you need.

GPU instances

Simple monthly plans

Compare plans and sizes

Free

One small sandbox to try CircuitNotion Cloud with your agent.

Free
  • 1 instance
  • 0.5 vCPU · 512 MB RAM
  • Public HTTPS URL
  • REST API + MCP access

Starter

For a production agent API and a couple of helpers.

RWF 10,000 / 30 days
  • 3 instances
  • 2 vCPU · 2 GB RAM pool
  • Deploy from Git
  • Private networking

Pro

Fleets of agents, tool servers and background workers.

RWF 30,000 / 30 days
  • 10 instances
  • 6 vCPU · 8 GB RAM pool
  • Deploy from Git
  • Priority support

Scale

Heavy multi-agent systems and data pipelines.

RWF 75,000 / 30 days
  • 25 instances
  • 16 vCPU · 24 GB RAM pool
  • xlarge instances
  • Dedicated onboarding

Questions

What is CircuitNotion Cloud?
A hosting platform for Linux servers (VPS) and container apps, designed so AI agents can manage them: everything in the dashboard is also available over a REST API and an MCP server.
What does agent-friendly mean here?
Agents are the primary users. They get docs in formats models read (/llms.txt and OpenAPI), JSON errors with a code and a hint, retry-safe creates, and API keys scoped to exactly what each agent should touch. More for agent builders.
Do I need an AI agent to use it?
No. The dashboard does everything: create instances, open a terminal in the browser, edit files, read logs and change settings.
How do I pay?
With MoMo or card through CircuitNotion Pay. Each payment keeps a plan active for 30 days. Nothing renews automatically; pay again from the billing page to continue.
What happens when my plan ends?
Your instances keep running for a 72-hour grace period. After that the account moves to the free plan and any instances beyond it are stopped, not deleted, so you can pay again and start them.
Do you have GPUs?
GPU instances with dedicated NVIDIA GPUs are managed with the same API. Join the GPU waitlist to tell us which GPUs, how many and when.
Can I run Docker or my own image?
Yes. Deploy any public Docker image as an app, build from a git repository's Dockerfile on paid plans, or use the Ubuntu + Docker template where the server supports it.

Give your agent a server

Start on the free plan. Upgrade when your agents need more instances, CPU or memory.

Start free with CircuitNotion