Message your local AI from anywhere.

Self-hosted relay + CLI that bridges your phone to a local ollama agent. Zero cloud. Zero API keys. Nothing ever leaves your network.

coming soon — get notified

The problem

Every "talk to your AI from your phone" setup routes your messages through someone else's server — Anthropic, OpenAI, a third-party relay. If you care about local-first, that's the opposite of what you want.

And if you're running a long-running local agent — a cron job that summarizes your day, a shell monitor, a writing assistant — there's no clean way to text it from the train.

The shape

[phone] ──POST──▶ [relay on your box] ◀──poll── [CLI] ──▶ [ollama]
                         │                       │
                         └────── reply ──────────┘

Four hundred lines of Node.js. Two files. Zero dependencies beyond the standard library. Your phone → your relay on hardware you own → your local ollama → back to your phone. Every hop stays on your network.

What's in the box

• relay-server.js — ~120 lines. Node HTTP server, four endpoints, in-memory by default.
• ollama-relay.js — ~150 lines. CLI bridge that polls the relay, calls ollama, posts back. Rolling conversation history.
• Systemd units — copy-paste ready for Debian/Ubuntu/Arch.
• Android shortcut recipe — a free FOSS app and five clicks puts the agent on your home screen.
• README — real configuration examples, security notes, the whole architecture in under ten pages.

Who it's for

• People running ollama locally who want to reach their agent from a phone.
• Self-hosters allergic to third-party APIs and SaaS middlemen.
• Developers building local-first agent workflows with gemma, llama, qwen, or any other ollama model.
• Anyone who wants a chat interface for a long-running cron job or shell script on their home box.

Honest limitations

• Home-LAN first. If you expose the relay to the internet, put it behind a reverse proxy with TLS.
• In-memory by default. Restart = message history gone. SQLite persistence is a 20-line patch.
• Auth is a single shared key. Good enough for your LAN. Bring your own TLS + stronger auth if you need more.
• Not a chat UI. It's the plumbing a chat UI would need.