Introducing SSH access for Runta runtimes — RUNTA

Start free trial

All BlogWhy do we give agents a shell?

Agents need persistent environments for long-running and recurring work, with the ability to scale beyond a single local computer.

Many agents already know how to use SSH, making it a natural interface for remote computers—no custom integration or agent-specific API required. With SSH access, an agent can use a Runta runtime like a cloud computer.

ssh my-agent

But giving an agent a shell raises an obvious question: does it also give the agent access to every credential and network resource on that computer?

With Runta, it does not. Agents can use authorized services without seeing the underlying credentials, while egress policies restrict which external destinations the runtime can reach.

Demo

Getting started

Prerequisites

Linux

npm install -g @runta/runta-cli<br>npx skills add https://runta.com/docs

macOS (Apple Silicon)

brew install runta-dev/tap/runta<br>npx skills add https://runta.com/docs

For agents

Give your agent the following instructions:

Set up a Runta runtime named my-agent and make it accessible over SSH.

Start by running runta login. Keep the command running while it waits for browser authorization. If it displays an authorization URL, show the URL to me so I can open it, then continue monitoring the same process until it reports that authentication is complete. Do not ask me to provide credentials, tokens, or confirmation that I have finished.

After login succeeds:<br>1. Run runta run --name my-agent --cpus 2 --memory 2048.<br>2. Run runta ps -a repeatedly until my-agent has finished creating and is ready. Wait at least 200 milliseconds between attempts.<br>3. Run runta ssh setup-alias my-agent. This command should create and register the local Runta device key automatically.<br>4. Run ssh my-agent to verify the connection.

If any step fails, inspect the error and report the exact blocker. Never display private keys, credentials, or authentication tokens.

For humans

# Log in to Runta with the CLI<br>runta login

# Start a new Runta runtime<br>runta run --name my-agent --cpus 2 --memory 2048

# Wait until the runtime is running<br>runta ps -a

# Inject the alias into ~/.ssh/config<br>runta ssh setup-alias my-agent

# Log in over SSH<br>ssh my-agent

Then, simply connect your agent to a Runta runtime over SSH and let it get to work.

About

Social