Skip to main content

Autonomous Intern

Autonomous Intern is an always-on AI device that handles work over messaging apps — email triage, scheduling, research, finance ops, and content drafting. It runs locally on a Raspberry Pi 5 powered by OpenClaw, the Hermes Agent, and Supermemory. Tasks come in over Telegram, Slack, or Discord — pick one of these as your first channel during initial pairing. WhatsApp can be added as an additional channel after the device is paired (it requires a QR pairing flow that the captive-portal wizard can't run). Email triage above means the Intern reads and triages your inbox as a task — email is not itself a transport for sending tasks to the device.

Product page: autonomous.ai/intern

These docs are for developers — people setting up an Intern in production, configuring Supermemory, or building on top of the Intern Developer Edition SDK.

What you can do with these docs​

  1. Setup Intern — provision a fresh device, pair it over Wi-Fi, and connect it to Telegram, Slack, or Discord (the only channels supported by the initial setup wizard — WhatsApp is added later as a second channel, see Channels → Adding more channels) and your LLM provider.
  2. Skills — how the device picks up role-specific skill bundles from GitHub on first session, the catalog of roles, and how the auto-update cronjob keeps them current.
  3. Supermemory on Intern — understand how the per-device Supermemory key is allocated, where it lives in config, and how to verify it's running.
  4. Setup Hermes — swap the default OpenClaw runtime for the Hermes Agent over SSH using the exact commands intern-server runs internally.
  5. Developer Edition — the Dev Edition ships with extra mic/speaker hardware and SSH access. Use the Python SDK to record audio, drive the speaker, and deploy code to the device.

Architecture at a glance​

Raspberry Pi 5 (Bookworm, ARM64)
├── intern-server         Go HTTP API on :5000 — setup, channels, LED, OTA
├── bootstrap-server      Go OTA worker — polls metadata.json every 5 min
├── openclaw.service      Node.js gateway — LLM + skill registry + channels
├── nginx                 Reverse proxy + captive portal SPA on :80
└── hostapd + dnsmasq     Wi-Fi AP "Intern-XXXX" for first-time pairing

Persistent state lives in /root/config/config.json (LLM key, channel tokens, Supermemory key/tag) and /root/.openclaw/openclaw.json (OpenClaw plugin config). Everything runs as root via systemd.

Hardware requirements​

  • Raspberry Pi 5 (4 GB RAM minimum) — Pi 4 and earlier are not supported
  • 64 GB microSD or onboard storage
  • USB-C power (5V / 5A recommended)
  • Wi-Fi 5 + Bluetooth 5.0 (built-in)
  • Developer Edition adds: external mic and speaker, exposed SSH

Where to start​

You want to…Go to
Flash a new device and get it talking on TelegramQuick Start
Switch a paired device to a new Wi-Fi networkPairing & Wi-Fi
See what role-specific skills ship with a paired InternSkills → Role Catalog
Change a device's role or force a skills updateManaging Skills
Understand why Supermemory is "installed: false"Supermemory Troubleshooting
Swap the agent runtime to Hermes by handHermes → Manual Install
Record audio from the Dev Edition micAudio SDK
SSH in and run a command remotelySSH Access