ɳSelfɳSELFCLI
DocsComparePricingChangelogBlogɳCloud
17★
All posts

Running your own AI assistant for $19/year

March 8, 2026·nSelf Team·3 min read
productaiplugins

Cloud AI assistants charge per seat, per message, or per month. Your data travels to their servers, gets processed, and comes back. You pay twice: once in dollars, once in privacy.

We built an alternative. ɳClaw is a self-hosted AI assistant that runs entirely on your own server. The plugin license costs $19/year. You bring your own LLM API keys and pay only for the tokens you actually use.

What ɳClaw does

ɳClaw is an ɳSelf plugin that adds a full AI agent runtime to your backend. It includes:

  • A ReAct reasoning loop that breaks complex requests into steps: reason about what it knows, decide which tool to call, execute, observe the result, repeat
  • Tool auto-discovery across your installed plugins. Install the analytics plugin and ɳClaw can query metrics. Install nself-google and it can read your calendar
  • Multi-provider support for OpenAI, Anthropic, Google, or a local Ollama instance
  • Persistent memory stored in your own Postgres database
  • Open-source client apps for iOS, Android, macOS, and web

Everything runs on your server. Conversations, memory, tool calls, and reasoning traces stay in your Postgres database. The only external call is to the LLM provider you choose.

The cost breakdown

Here is what a typical personal or small-team setup costs per year:

ItemCloud AI serviceɳClaw on ɳSelf
Software license$240-600/yr$19/yr (Max tier)
ServerIncluded$4-8/mo (Hetzner CX23)
LLM tokensIncluded$5-15/mo (your own keys)
Annual total$240-600$67-195

The Max tier at $19/year includes ɳClaw plus the entire AI suite: nself-ai (the inference engine), nself-mux (email routing), nself-voice, and nself-browser. All five plugins for the price of one month of most cloud alternatives.

Setting it up

If you already have an ɳSelf server running, adding ɳClaw takes about two minutes:

nself license set nself_pro_...
nself plugin install ai mux claw

Add your LLM provider to .env:

PLUGIN_CLAW_PROVIDER=anthropic
PLUGIN_CLAW_API_KEY=sk-ant-...

Then rebuild and restart:

nself build && nself restart

That is it. Your AI assistant is running at https://claw.your-domain.com. Connect the open-source client apps or use the web interface.

Tool auto-discovery

This is the part that makes ɳClaw different from wrapping an LLM in a chat UI.

Every ɳSelf plugin that supports ɳClaw exposes a /internal/tools endpoint. When ɳClaw starts, it polls all installed plugins and builds a tool registry automatically. No manual tool definitions. No JSON schemas to write.

Install nself-google and ɳClaw can read your calendar, search your Drive, and draft emails. Install nself-stripe and it can look up customer subscriptions. Install nself-cms and it can query and update your content.

The tools compose naturally. Ask ɳClaw to "find my meetings tomorrow and draft a prep email for each one" and it will call the calendar tool, loop over results, and call the email tool for each meeting. All in one request, all on your server.

Privacy by default

With hosted AI services, every prompt goes to their server. Every document you paste, every question you ask, every file you upload becomes part of their processing pipeline. For teams with customer data, financial records, or internal communications, that is a real problem.

ɳClaw never sends your data anywhere except the LLM provider you chose. And even that call contains only the current conversation context, not your entire database. Tool results stay local. Memory stays local. Reasoning traces stay local.

If you run a local model through Ollama, nothing leaves your network at all.

The tiers

ɳSelf plugins use a simple tier system:

  • Free ($0): 16 MIT-licensed plugins. Everything you need for a basic backend
  • Pro ($9/year): 55+ plugins including chat, notifications, cron, CMS, Stripe, and more
  • Max ($19/year): Pro plus the AI suite. ɳClaw, nself-ai, nself-mux, nself-voice, nself-browser

These are launch prices. They will go up for new subscribers as the ecosystem matures, but existing subscribers keep their rate.

Try it

Start with a fresh ɳSelf install or add ɳClaw to your existing server:

brew install nself-org/nself/nself
nself init my-project && cd my-project
nself license set nself_pro_...
nself plugin install ai mux claw
nself build && nself start

Full docs at docs.nself.org. Source code for the client apps at github.com/nself-org/claw.