OpenClaw — Build Your Personal AI Agent

Build Your Own Powerful Personal AI Agent

Use OpenClaw + Telegram to deploy your AI agent in the cloud or locally — powered by free-tier LLMs. No China account needed.

$0 Total Cost

Stay on free tiers

70M+ Free Tokens

Alibaba new users

4 Days

1 hour per day

Day 1

Cloud Deployment on Alibaba Cloud + Telegram

View Details

Goal Have a fully working OpenClaw agent running 24/7 in the cloud, chatting via Telegram.

1Sign up / log in at alibabacloud.com (use non-China region).
2Follow the official one-click tutorial: purchase the OpenClaw pre-installed image.
3In the console → Application Details → Execute the three commands.
4Telegram integration: Create bot with @BotFather, paste token into OpenClaw config.
5Live test: Message your bot → "Tell me about yourself".

HomeworkKeep the instance running overnight and note your bot username.

Day 2

Local Self-Hosted Setup + Free Cloud LLMs + Telegram

View Details

Goal Run OpenClaw on your own laptop with zero ongoing cost, using ultra-fast free cloud models.

1Recap: Review Day 1 cloud instance (you can stop it to save credits).
2Install locally: curl -fsSL https://openclaw.ai/install.sh | bash, then run openclaw onboard.
3Add free LLM — Groq recommended: get free API key at console.groq.com/keys.
4Test & compare: Chat with your local agent vs. the Alibaba cloud one.

HomeworkAdd a second free fallback model (e.g. Gemini) in ~/.openclaw/openclaw.json.

Day 3

NVIDIA Free Kimi K2.5 + Smart Fallbacks + Advanced Config

View Details

Goal Add the powerful free Kimi K2.5 and configure automatic fallbacks.

1Recap: Test Day 2 local setup.
2NVIDIA free Kimi setup: Get API key at build.nvidia.com/settings/api-keys, add as custom provider, set model moonshotai/kimi-k2.5.
3Configure smart fallbacks: Kimi primary → Groq fallback → Gemini fallback.
4Run openclaw restart and test with heavy tasks.
5Live demo: Watch the agent automatically switch models when Kimi rate-limits.

HomeworkTry adding one more NVIDIA model (e.g. GLM-4.7 or Qwen3).

Day 4

Skills, Memory, Agents & Next-Level Projects

View Details

Goal Turn your OpenClaw into a true personal agent that remembers you, runs skills, and does real work.

1Recap: Quick poll — who is running local vs. cloud?
2Persistent memory: Teach it about yourself. Explore built-in skills.
3Community skills: Browse clawhub.com → install 2–3 skills. Create your first custom skill.
4Capstone project: Group challenge — "Make your agent book a fake meeting and summarize the web."
5Final Q&A, share your bot usernames, and next steps.

HomeworkUse your agent daily for 7 days and report back one cool thing it did.

Build Your Own Powerful Personal AI Agent