Run code for AI agents
Run code and tests in an isolated sandbox. Start compute when you need it and pay for its runtime.
What it's for
Spin up an isolated code sandbox — CPU or GPU — on demand.
Run a test suite, evaluate a snippet, or train something small.
State persists across commands in the same session until closed.
Billed per second of uptime — only while the sandbox is open.
How to ask your agent
Connect Vaaya, then just say it in plain language. For example:
❯Use Vaaya to spin up a sandbox and run this repo's tests on Node 22.
❯Use Vaaya to open a GPU sandbox and run this training script.
What you get back
{
"session_id": "e2b_7a31f0",
"stdout": "PASS tests/sum.test.ts\n ✓ adds numbers (2ms)\n\nTest Suites: 1 passed, 1 total",
"stderr": "",
"returncode": 0
}
Pricing
from ~1¢ for a single quick run; metered sandboxes ~5¢/vcpu-hour while open. Charges are shown in cents and recorded in your transaction history. Set a maximum cost for each call. Failed calls are not charged.
Try it with your agent.
Connect Vaaya and describe what you want to do. Your agent can find a suitable service and call it within your budget.
npx @vaaya/mcp install