💻 Give an agent a machine to work on
Stack: Runner: any machine (laptop, build server, VPC box) · Connection: outbound only · Surface: commands + installed runtimes + filesystem · Output: work done on that box
Who it’s for: anyone whose agent needs to reach something a hosted agent can’t: internal services, on-prem data, private infrastructure. What it does: a small runner process you install on any machine gives agents a place to execute, run commands, use the runtimes already installed, touch that box’s files and network. Code, data, and credentials stay put; only instructions cross the wire.
Build it
-
Install and start it on the machine:
npx @almyty/auth login npx @almyty/runner start --name build-serverIt connects out (no inbound ports) and starts heartbeating.

-
It reports what it can do. Open the runner and it shows the runtimes and tools it detected, Node, Python, Docker, Git, so an agent knows what’s available.

-
Point an agent at it. The agent executes on that machine as a step: run a test suite, query a DB only reachable inside the network, or process files that can’t leave.
Make it yours
- Driving coding-agent CLIs is one capability built on this, not the point of the runner.
- Reach a database that only listens on
localhostinside a private subnet. - Keep regulated data on-prem while the control plane stays hosted, or self-host all of it.
Uses: Platform · Runner · On-prem