Checking platform status…

Run AI-generated code. Safely.

Vertal Run gives AI agents and developers isolated sandboxes to execute code: one API, an SDK, and an MCP server — with keys, quotas and usage metering built in.

Open the playground Get your API key View on GitHub

Platform

Everything an agent needs to act — nothing it can break.

Each execution runs in a fresh sandbox that is destroyed afterwards. Built from scratch by KUBO Protocol, with no third-party sandbox API in the loop.

⛨

Isolated by default

No network, read-only filesystem, unprivileged user, CPU, memory and process limits. Timeouts kill the sandbox.

⚡

Fast

About a quarter of a second per execution end to end. Python, JavaScript and Bash today.

✦

Made for AI agents

A native MCP server exposes run_code to KUBO IA, Claude and any MCP client — with the same quotas as the API.

◎

Usage-based

API keys, plans, daily quotas, rate limits and per-tenant metering. Every request carries a request ID.

Playground

Run code on the live platform.

Paste your API key, write code and execute it in a real sandbox. Your key stays in this browser tab only.

Code

import statistics

prices = [42.0, 39.5, 44.2, 47.9, 45.1]
print("mean:", round(statistics.mean(prices), 2))
print("stdev:", round(statistics.stdev(prices), 2))

Language

Timeout (s)

API key

Output

# Output appears here.

Pricing

Start free. Scale to millions of executions.

Limits below are enforced by the platform today.

Free

For prototypes

100 executions / day
10 requests / minute
10 s max run time
256 MB sandboxes

Start building

Most popularPro

$29 / month

For products in production

10,000 executions / day
120 requests / minute
60 s max run time
1 GB sandboxes

Get Pro

Enterprise

Custom

For platforms and agents at scale

1,000,000 executions / day
1,200 requests / minute
300 s max run time
4 GB sandboxes

Talk to us

Docs

Integrate in minutes.

The essentials are below. The full guide is in the documentation.

Get an API key

Keys look like kubo_sk_…, are shown once and stored hashed.

Call the API

POST /api/v1/executions with your code. You get stdout, stderr, exit_code, duration and usage.

Connect your agent

Add the MCP server and your agent gets a run_code tool.

Track usage

GET /api/v1/usage returns today's executions, sandbox seconds and plan limits.

# HTTP
curl -X POST https://your-domain/api/v1/executions \
  -H "Authorization: Bearer $KUBO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"language":"python","code":"print(6*7)"}'

# Python SDK
from vertal2kubo.client import KuboClient
client = KuboClient("https://your-domain/api", api_key)
print(client.run("print(6*7)")["stdout"])  # 42

# MCP (Claude, KUBO IA, IDEs)
{"mcpServers": {"kubo": {"command": "python",
  "args": ["-m", "vertal2kubo.mcp"]}}}