How it works

Your stack

Ollama, Qwen and OpenClaw in one guided install—no scattered tutorial glue code.

Ollama Local runtime on your machine, installed during setup.

Qwen3.5:9b Model pulled through Ollama for your private assistant.

OpenClaw Agent layer that executes your workflows and automations.

Hardware requirements: 8GB VRAM.

Web search

Get a Tavily API key

Tavily supplies web search for tools that need fresh results from the internet. The free tier includes up to about a thousand searches per month.

Create an account at tavily.com.
Open the Tavily dashboard → API Keys.
Create or copy your key (it usually starts with tvly-).
Add it to your local .env file as TAVILY_API_KEY=tvly-…. The installer can merge this into your OpenClaw configuration.

Gmail bridge

Maton API key

Maton provides a Gmail OAuth proxy so automations can work with email without you handing raw passwords to every script.

Sign in at maton.ai and open your API Keys (or account) page.
Create or copy an API key.
Put it in your local .env as MATON_API_KEY=….

Telegram bot and your user ID

Anyone who finds your bot’s username can try to message it, so you pair a bot token with an allowlist: only your numeric Telegram user ID can drive your local assistant.

Create a Telegram bot

In Telegram, open a chat with @BotFather.
Send /newbot.
Pick a display name and a unique username that ends with bot.
BotFather replies with an HTTP API token that looks like <bot-id>:<secret>.
Save it in your local .env as TELEGRAM=<bot-id>:<secret>.

Open a direct chat with your new bot and send /start once so Telegram allows the bot to deliver replies to you. You can later use BotFather’s /revoke if you need to rotate the token.

Find your Telegram user ID

Search for raw_data_bot (@raw_data_bot).
Start the bot and send a message if it asks you to.
In the reply, copy your numeric user ID (digits only).
Add TELEGRAM_ALLOW_FROM= that number in .env so only your account can use the bot with your setup.