PUDA CLI and Skills
Install the PUDA CLI and PUDA skills in the shell environment used by your AI agent. This lets Cursor, Claude, Hermes, or any shell-capable agent discover machines, run PUDA commands, and follow PUDA workflows.
It works best on Linux, where the agent can install the PUDA CLI, inspect project files, run commands, and operate through the same terminal surface that a human would use.
Copy this prompt into your AI agent to run the full setup in one go:
Or follow the steps below.
Install prerequisites
Copy this prompt into your AI agent:
Before installing the PUDA CLI, install the latest available versions of Python 3, pip, and npx for this system.
After installation, verify that python3, pip, and npx are available by checking their installed versions.
Install the PUDA CLI
Copy this prompt into your AI agent:
Install the latest PUDA CLI 0.1.x from https://github.com/PUDAP/puda/releases and add it to my PATH so the puda command is available globally.
After installation, verify that the CLI reports a 0.1.x version.
Install PUDA skills
Copy this prompt into your AI agent:
Install PUDA skills using `puda skills install`.
Log in
Copy this prompt into your AI agent, replacing <username> with your PUDA username:
Log in to PUDA using the username <username>.
Configure NATS
Copy this prompt into your AI agent, replacing <nats_servers> with the server addresses from your NATS setup:
Inspect PUDA profiles and configure a named profile for the NATS server URL or comma-separated NATS cluster URLs: <nats_servers>. Select it as the active profile. If it uses TLS and E2EE, ask me for the local CA and lab key file paths.
Use puda profile for NATS server connections; puda config manages user identity only. Plaintext profiles use nats:// URLs. Encrypted profiles use tls:// URLs with both --ca and --key. See profile migration and examples.
Verify the setup
Copy this prompt into your AI agent:
List the current available PUDA machines.
The machines with edge services running should show up there.
Use puda machine info <machine_id> to inspect a machine's advertised tlm_streams, then puda machine watch <advertised-subject> --timeout 15 to verify live samples. Watch takes full NATS subjects, not machine IDs. See watch subject examples.
Next, build machine skills so your AI agent knows how to command each machine safely and consistently.