Getting started

Quick start

Five minutes from zero to a running OpenClaw agent you can talk to in Telegram. Assumes you already have EXMER deployed at your own domain (if not, see Installation first).

1. Open the bot in Telegram

Find @EXMERBot in Telegram (or whatever name you gave your deployment), tap Start, then tap the Open EXMER button to launch the Mini App.

Auth. EXMER uses Telegram's initData signature to verify your identity. You don't enter a password — the bot knows who you are. Your Telegram user ID must be in the deployment's ADMIN_USER_IDS environment variable for the first login to work.

2. Add your server

On the Dashboard, tap + Add and fill in:

Name — any label you'll recognize (e.g. prod-main).
Host — the server's IP or domain.
Port — SSH port, default 22.
Username — root or a user with sudo.
Auth type — password or private key. The key is encrypted with AES-256-GCM before hitting the DB.

EXMER runs a test connection, detects whether OpenClaw is already installed, and you're done.

3. Install OpenClaw (if needed)

If OpenClaw isn't on the server yet, tap Install OpenClaw. An eight-step wizard runs:

Check prerequisites (curl, build-essential, etc.)
Install Node.js 22 via NodeSource
Install OpenClaw CLI + grammy (for Telegram channels)
Run openclaw onboard to create the initial config
Configure gateway (bind, port, auth mode)
Set up systemd unit and start the gateway
Verify openclaw health returns green
Create the first default agent

Each step reports progress in real time. If something fails, you see the exact error and can retry just that step.

4. Configure an agent

Tap Agents → pick your agent → expand the sections:

Identity — display name and emoji
IDENTITY.md — character description the LLM reads
SOUL.md — behavioral rules and values
Model — which Claude / OpenAI model to use + thinking level
Skills — enable bundled OpenClaw skills, add custom ones
API Keys — paste your Anthropic / OpenAI / Voyage keys
Telegram — bind the agent to a Telegram channel

5. (Optional) Give the agent a knowledge base

Tap 🧠 RAG / Memory from the Agent page to open the RAG manager. Pick a category — local Ollama (free) or cloud (Voyage / OpenAI, paid) — enter an API key if needed, upload your documents, run Reindex, then Enable for agent.

See the RAG memory page for details.

6. Say hello

Open the Telegram channel your agent is bound to and send a message. The agent responds through OpenClaw using the model and knowledge you configured.

Nothing is hidden. Every setting you touched in the UI maps 1:1 to files on the server under ~/.openclaw/. You can always SSH in and inspect them. The Terminal page is there for exactly that.