OpenClaw<\/a> What is it<\/strong><\/p>\nOpenClaw is essentially a locally run AI gateway<\/p>\n
Its structure is simple:<\/p>\n
{\\cHFFFFFF}{\\cH00FFFF} Your computer runs a Gateway process {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF} Gateway {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF} {\\cHFFFF00} You send a Gateway to AI \\AI \\AI \\AIgram<\/p>\n
Why would you do that<\/p>\n
Because most of the AI services (Claude, GPT) do not directly provide Telegram robots, you use either the web version or the App<\/p>\n
But OpenClaw allows you to talk to AI in any chat software, and all memories and configurations are on your own computer (~\/.openclaw\/), without risk of privacy disclosure<\/p>\n
Second, get ready for the prep<\/strong><\/p>\n1. Hardware and system requirements<\/strong><\/p>\nMac Users: Best Saved, Native Support MacOSWindows Users: WSL2 (Windows Subsystem for Linux) strongly recommends Ubuntu 22.04Linux Users: Directly<\/p>\n
If you want to hang up 24 hours, suggest:<\/p>\n
\n- Mac Mini (most electricity, official recommendation)<\/li>\n
- Cloud Server VPS (Digital Ocean 12 US$\/month, Hostinger is also available and domestic)<\/li>\n<\/ul>\n
2. LARGE MODEL API READINESS<\/strong><\/p>\nOpenClaw supports almost all mainstream models, but you need to be ready in advance<\/p>\n
Comparison of recommended programmes:<\/p>\n
ChatGPT Plus Subscription<\/p>\n
$20\/MONTH, OFFICIAL WHITE LIST SUPPORTED, MOST ECONOMICAL<\/p>\n
Coding Plan<\/p>\n
Qwen3-Max\/kimi k2.5<\/p>\n
OpenRouter (\u2b50\u2b50\u2b50\u2b50)<\/p>\n
BASED ON VOLUME, APPROXIMATELY $3-15\/DAY, MULTI-MODEL SWITCHING, SUITABLE FOR TESTING<\/p>\n
AI\/ML API (\u2b50\u2b50\u2b50\u2b50)<\/p>\n
Pays in volume, supports multiple models, OpenClaw official integrated partners<\/p>\n
Kimi K2.5 (\u2b50\u2b50\u2b50\u2b50)<\/p>\n
Pays by volume, and the dark side of the moon actively supports OpenClaw<\/p>\n
Transit stations (\u2b50\u2b50\u2b50)<\/p>\n
Cheap but unstable, high risk<\/p>\n
Ollama Local Model<\/p>\n
It's completely free, cost-free, but not effective, suitable for learning<\/p>\n
My suggestions:<\/p>\n
\n- Start with OpenRouter or AI\/ML API, pay in volume, don't waste<\/li>\n
- make sure it's a long-term one<\/li>\n
- Don't get up to the stop. You don't know if it's a model or a configuration<\/li>\n<\/ul>\n
3. Wallbreaking environment (important)<\/strong><\/p>\nTelegram and Discord both need to cross the wall to connect<\/p>\n
You need:<\/p>\n
\n- Clash Verge \/ Clash for Windows \/ Surge<\/li>\n
- A stable airport node<\/li>\n
- TURN ON TUN MODE (MANY PEOPLE FORGET, DON'T OPEN TUN MODE, TERMINAL CANNOT WALK AGENT)<\/li>\n<\/ul>\n
Validation method:<\/p>\n
i can't believe it<\/p>\n
If returned 200, indicate that your terminal can access Telegram API<\/p>\n
The official installation will follow you<\/strong><\/p>\nStep 1: Install Homebrew and Node.js<\/p>\n
Open Terminal with Homebrew:<\/p>\n
\/bin\/bash -c \"$(curl-fsSL)\"<\/p>\n
https:\/\/raw.githubusercontent.com\/Homebrew\/install\/HEAD\/install.sh<\/p>\n
Node.js:<\/p>\n
i'm going to tell you something<\/p>\n
Validation installation:<\/p>\n
node \u2013version # should show v22.x.x or higher<\/p>\n
npm -version<\/p>\n
Important: OpenClaw requires Node.js version 22, less than this version will be wrong<\/p>\n
Step 2: Install OpenClaw<\/p>\n
Recommended mode (official guide):<\/p>\n
quil-fssL bash https:\/\/openclaw.ai\/install.sh<\/p>\n
This script will automatically help you:<\/p>\n
\n- Install OpenClaw CLI tool<\/li>\n
- create profile directory (~\/.openclaw\/)<\/li>\n
- Set Environment Variable<\/li>\n<\/ul>\n
Once installed, run:<\/p>\n
openclaw -version<\/p>\n
Seeing the version number means it's loaded<\/p>\n
Step 3: Start Configuration Wizard<\/p>\n
This is the most critical step:<\/p>\n
this is an openclaw on board-install-daemon<\/p>\n
\u2013install-daemon parameters allow OpenClaw to continue in the backstage, even if the computer is restarted<\/p>\n
And then you're going to go into an interactive wizard, in a few steps:<\/p>\n
IV. The configuration process is where most people are stuck<\/strong><\/p>\n4.1 SELECTION OF AI MODEL PROVIDER<\/strong><\/p>\nThe wizard lists:<\/p>\n
\n- OpenAI (GPT)<\/li>\n
- i'm sorry<\/li>\n
- ….<\/li>\n
- Skip<\/li>\n<\/ul>\n
If you have an official account number: select the corresponding one, then jump through the browser<\/p>\n
If you use a third party API (OpenRouter, Transit etc.): Choose Skip<\/p>\n
After skipping, the wizard will continue<\/p>\n
4.2 Select Chat Platform<\/strong><\/p>\nThe wizard will ask which platform you want to connect to:<\/p>\n
\u2022 Telegram (Bot API)<\/p>\n
Whatsapp (QR code)<\/p>\n
\u00b0 Discord (Bot API)<\/p>\n
Skip<\/p>\n
Why recommend Telegram:<\/p>\n
\n- The simplest configuration (three minutes done)<\/li>\n
- Bot creation without review<\/li>\n
- Support long message and file transfer<\/li>\n
- Full Bot Management Panel<\/li>\n<\/ul>\n
Discord for whom:<\/p>\n
\n- There already is a team of Discord servers<\/li>\n
- It takes a lot of people to work with<\/li>\n
- Configure time approximately 10 minutes [7]<\/li>\n<\/ul>\n
Here we go first, Telegram<\/p>\n
4.3 Create Telegram Bot<\/strong><\/p>\nOpen Telegram, search @BotFather<\/p>\n
send \/start start dialogue<\/p>\n
Then send in order:<\/p>\n
\n- \/newbot<\/li>\n
- Enter your Bot display name (e. g. my AI assistant)<\/li>\n
- Enter Bot username (must end with _bot, e. g. ai_bot)<\/li>\n<\/ol>\n
After success, BotFather returns a Token format similar to:<\/p>\n
1234567890: ABCdefGHIjklMNOpqrsTUVwxyz<\/p>\n
Copy this Token<\/strong>Back to the terminal. Paste in<\/p>\n4.4 Completion of initial configuration<\/strong><\/p>\nThe wizard will ask if you want to install Skills, for example:<\/p>\n
\n- Web Search<\/li>\n
- Mail Management<\/li>\n
- Organisation<\/li>\n
- File Operations<\/li>\n<\/ul>\n
The new guy suggested we skip first, wait for the basics to go through<\/p>\n
After the configuration is completed, the Wizard will automatically start Gateway:<\/p>\n
\u2022 Gateway started on http:\/\/127.0.0.1:18789<\/p>\n
Telegram channel conned<\/p>\n
Ready to chat!<\/p>\n
CONFIGURE THIRD-PARTY AI MODELS<\/strong><\/p>\nIf you choose Skip in 4.1, now you need a manual configuration model (the following steps strongly recommend cursor and so on to be an assistant<\/p>\n
Method 1: Through Web UI Configuration (Recommended)<\/p>\n
Open browser, access:<\/p>\n
http:\/\/127.0.0.1:18789<\/p>\n
This is OpenClaw's control panel<\/p>\n
Click Settlings \u2192Models \u2192Add Projecter<\/p>\n
FILL IN YOUR API MESSAGE:<\/p>\n
\n- Provider Name: Custom (e. g. openrouter)<\/li>\n
- Base URL: Your API address<\/li>\n
- Your key<\/li>\n
- Model ID: Specific model name (e.g. claude-opus-4)<\/li>\n<\/ul>\n
After saving, texting test for Bot in Telegram<\/p>\n
Method 2: Directly edit configuration files<\/p>\n
Open profile:<\/p>\n
~\/.openclaw\/openclaw.json<\/p>\n
found models.providers section, add:<\/p>\n
{\u201cModels\u201d: {\u201cproviders\u201d: {\u201copenrouter\u201d: {\u201cbaseUrl\u201d: \u201chttps:\/\/openrouter.ai\/api\/v1\u201d, \u201capiKey\u201d: \u201cyour-api-key-here\u201d, \u201capi\u201d: \u201copenai-complements\u201d, \u201cmodels\u201d: [{id\u201d: \u201canthropic\/claude-opus-4\u201d, \u201cname\u201d: \u201cClaude Opus 4\u201d, \u201creasing\u201d: True}}, \u201cagents\u201d:}, {\u201cdefaults\u201d: {model: {primary\u201d: {openrouter\/antrope-opus-4}}<\/p>\n
Start again after saving<\/p>\n
openclaw present<\/p>\n
METHOD 3: USE AI\/ML API (OFFICIAL INTEGRATION)<\/p>\n
IF YOU USE AI\/ML API, YOU CAN USE THEIR SPECIAL VERSIONS DIRECTLY WHEN INSTALLED:<\/p>\n
@latest<\/p>\n
This version will automatically configure the AI\/ML API Skills<\/p>\n
VI. CONFIGURE THE NETWORK AGENT SO THAT YOUR AI CAN ACCESS THE NETWORK<\/strong><\/p>\nBot didn't get back<\/p>\n
Confirm your proxy port<\/p>\n
Open Clash Verge, see:<\/p>\n
\n- HTTP PORT (GENERALLY 7890)<\/li>\n
- SOCKS5 PORT (GENERALLY 7891)<\/li>\n<\/ul>\n
Set Environment Variable<\/p>\n
edit shell profile:<\/p>\n
# if with zsh (Mac Default)<\/p>\n
~ \/.zshrc<\/p>\n
# if use bash<\/p>\n
~\/.bahrc<\/p>\n
Add at the end of the document:<\/p>\n
export http_proxy=http:\/\/127.0.0.1:7890<\/p>\n
export https_proxy=http:\/\/127.0.0.1:7890<\/p>\n
for example, all_proxy=socks5:\/\/127.0.0.1:7891<\/p>\n
Save and reload:<\/p>\n
~\/.zshrc # or ~\/.bashrc<\/p>\n
Restart OpenClaw:<\/p>\n
openclaw present<\/p>\n
Validation agent:<\/p>\n
Message to Bot in Telegram: Search OpenClaw for updates<\/p>\n
If you can return the result, indicate that the agent's configuration is successful<\/p>\n
VII. PROGRESS CONFIGURATION MAKES AI REALLY USEFUL<\/strong><\/p>\n7.1 CONFIGURE PERSONALISED SOUL<\/strong><\/p>\nOpenClaw supports custom AI character and behaviour rules<\/p>\n
Edit files:<\/p>\n
nono~\/.openlaw\/workspace\/SOUL.md<\/p>\n
Write your requests, for example:<\/p>\n
# MY IAA ASSISTANT PERSONALISATION - CALL ME \"BOSS\" - RESPONDS STYLE: SIMPLICITY, NO BULLSHIT - EXPERTISE IN THE FIELD: PRODUCT DESIGN, TECHNICAL STRUCTURES, CONTENT CREATION - BAN: DON'T ACTIVE TALK, DON'T SAY \"SORRY\" THAT'S THE KIND OF TALK<\/p>\n
ONCE SAVED, AI WILL RETURN TO YOU WITH THIS SETUP<\/p>\n
7.2 Security configuration (this must be done)<\/strong><\/p>\nOpenClaw was warned by a safety researcher about the risks because of the overload<\/p>\n
Security measures that must be put in place:<\/p>\n
\n- Limit API spending caps on your API provider backstage (OpenRouter\/Claude etc.) to a daily\/monthly budget cap<\/li>\n
- enable sensitive operation to confirm edit configuration file, find tables section:<\/li>\n<\/ol>\n
{\u201ctools\u201d: {\u201cexec\u201d: {enabled\u201d: true, \u201capproval Required\u201d: true \/ \/ \/ need your confirmation before executing the command}, \u201cemail\u201d: {enabled\u201d: true, \u201capproval Required\u201d: true \/ \/ e-mail requires your confirmation }<\/p>\n
\n- OpenClaw is not exposed to the public web Gateway default listening 127.0.0.1:18789, only in-house access, and never change to 0.0.0.0<\/li>\n<\/ol>\n
VIII. CLASSING ISSUES<\/strong><\/p>\nQuestion 1: Bot doesn't reply<\/strong><\/p>\nQuest steps:<\/p>\n
# 1. Check whether Gateway is running openclaw status<\/p>\n
# 2. View logbooks<\/p>\n
# 3. Test model connection openclaw anent \u2013message \"test\" -thining high<\/p>\n
If logs see ECONNREFUSED or timeout, it's a web-based proxy problem<\/p>\n
If you see 401 Unauthorized, it's API Key error<\/p>\n
If you see arate limitt, it's your API level<\/p>\n
Question 2: Port occupied<\/strong><\/p>\nError: listen EADDRINUE: address already in use #18789<\/p>\n
Resolve:<\/p>\n
# PROCESS OF FINDING PORT OF OCCUPATION<\/p>\n
lsof-i:18789<\/p>\n
# KILL PROCESS (PID IS THE LAST AVAILABLE NUMBER)<\/p>\n
you know, kill-9 on the ped<\/p>\n
# Restart OpenClaw<\/p>\n
openclaw present<\/p>\n
<\/pid><\/p>\n
Question 3: Telegram pair failed<\/strong><\/p>\nIf Bot creates but does not match:<\/p>\n
# VIEW PAIRING<\/p>\n
openclaw calling list<\/p>\n
# MANUAL APPROVAL PAIR<\/p>\n
your Telegram User ID<\/p>\n
Your User ID can be retrieved from Telegram@userinfobot<\/p>\n
IX. COSTS AND POSITION RECOMMENDATIONS<\/strong><\/p>\nWe'll talk about the cost<\/p>\n
Minimum cost scenario<\/strong>(approximately 30 yuan\/month):<\/p>\n\n- Ollama Local Model<\/li>\n
- Telegram<\/li>\n
- My own computer is on 24 hours a day<\/li>\n
- Total: 20-30 yuan\/month<\/li>\n<\/ul>\n
Price-for-money scheme<\/strong>(approximately 150 yuan\/month):<\/p>\n\n- OpenRouter Pays by Volume (approximately $100\/month, light)<\/li>\n
- Telegram (free of charge)<\/li>\n
- CLOUD SERVER VPS (APPROXIMATELY 50 YUAN\/MONTH)<\/li>\n
- Total: 150 yuan\/month<\/li>\n<\/ul>\n
The Too programme<\/strong>(approximately $1,000 per month):<\/p>\n\n- direct to claude api<\/li>\n
- Mac Mini local running (one-time input of $3000)<\/li>\n
- Total: First month 4000+, follow-up 1000+\/month<\/li>\n<\/ul>\n
My advice: run for a week or two, make sure you really use it, and then think about promotion<\/p>\n
It's written at the end<\/strong><\/p>\nOpenClaw's core value is not another AI chat tool<\/p>\n
It's to shorten the process of \"open the browser to log in and copy the results\" to \"send a message at Telegram\"<\/p>\n
This is the productivity tool that super-individuals need: seamless, ready, continuous memory<\/p>\n
But it's not all-powerful either. Current problems include:<\/p>\n
\n- It's too fast to update<\/li>\n
- Skills, of mixed quality, some with security risks<\/li>\n
- Configure complexity is not very friendly to white<\/li>\n<\/ul>\n
But none of this prevents it from becoming the most promising open source project in 2026<\/p>\n
If you have problems, don't give up<\/p>\n
Claude Code or GitHub in OpenClaw<\/p>\n
THE TECHNOLOGY IS ALWAYS FOR PEOPLE<\/p>","protected":false},"excerpt":{"rendered":"
This is the first one of our openclaw teachings, where you can click first, and the next three-to-fives of the plan will be written, with a detailed use of middle-to-high-level, first, to find out what OpenClaw is, essentially, a locally run AI gateway, and its structure is simple: {\\cHFFFFFF}{\\cH00FFFF} Your computer runs a Gateway process {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF}{\\cH00FFFF} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF} Getway {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF}{\\cH00FF00} {\\cHFFFFFF} Getaway {\\cHFFFFFF}{\\cH00} {\\cHFFFFFF} {\\cHFFFFFF} Getway {\\cHFFFF00} Connect to your chat software (Telegram\/ Discord\/ Whatsapp) {\\cH0000000000} Connect to the AI {\\3030300000} {\\cH0000000000} You send a message from Telegram<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[149,144],"tags":[8229,8392],"collection":[8384],"class_list":["post-51300","post","type-post","status-publish","format-standard","hentry","category-jiaocheng","category-baike","tag-openclaw","collection-openclaw"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/51300","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=51300"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/51300\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=51300"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=51300"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=51300"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=51300"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"20da32e8j00tc4xk0047d000iw007kp\"](\"https:\/\/www.1ai.net\/wp-content\/uploads\/2026\/03\/20da32e8j00tc4xzk0047d000iw007kp.jpg\" "\\"20da32e8j00tc4xk0047d000iw007kp\\"")
<\/p>\n