
Preparing the next screen, account state, and workspace data.
This guide keeps the setup short and readable: create your API key, install OpenClaw on Node 22, paste one config file, restart the gateway, then optionally connect Telegram.
Use your own Aimesh endpoint here, not the upstream provider URL.
Use OpenClaw with the Responses API path to avoid the old completions mismatch.
Keep the 128k context profile so OpenClaw does not compact too aggressively.
Start in the dashboard first. OpenClaw should use your user API key from this site, not the upstream provider key.
OpenClaw requires Node 22.12+ or newer. If you install it under another Node version, the command may disappear after you switch versions.
Use one clean config file that points OpenClaw at this site's /api/v1 endpoint and keeps the model alias simple.
After the file is saved, validate the config, inspect the model status, then restart the gateway so the new provider settings are loaded.
If you want users to talk through Telegram, add a Telegram bot token, restart the gateway again, then send a message to the bot.
Start in the dashboard first. OpenClaw should use your user API key from this site, not the upstream provider key.
Use your own Aimesh endpoint here, not the upstream provider URL.
Your customers should use the user API key issued by this site. They should not use the upstream provider key directly.
Keep the 128k context profile so OpenClaw does not compact too aggressively.
OpenClaw requires Node 22.12+ or newer. If you install it under another Node version, the command may disappear after you switch versions.
source ~/.nvm/nvm.sh
nvm install 22
nvm use 22
SHARP_IGNORE_GLOBAL_LIBVIPS=1 npm install -g openclaw@latest
openclaw --versionIf openclaw is missing right after nvm use 22, reinstall it under Node 22. Global npm tools are tied to the active Node version.
Use one clean config file that points OpenClaw at this site's /api/v1 endpoint and keeps the model alias simple.
open "$HOME/.openclaw"{
"models": {
"mode": "merge",
"providers": {
"aimesh": {
"baseUrl": "https://aimesh.pro/api/v1",
"apiKey": "paste-your-aimesh-api-key-here",
"api": "openai-responses",
"models": [
{
"id": "gpt-5.4",
"name": "gpt-5.4 (Aimesh)",
"api": "openai-responses",
"reasoning": false,
"input": ["text"],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 128000,
"maxTokens": 32000
}
]
}
}
},
"agents": {
"defaults": {
"model": {
"primary": "aimesh/gpt-5.4"
},
"models": {
"aimesh/gpt-5.4": {
"alias": "gg"
}
},
"contextTokens": 128000
}
}
}After the file is saved, validate the config, inspect the model status, then restart the gateway so the new provider settings are loaded.
source ~/.nvm/nvm.sh
nvm use 22
openclaw config validate
openclaw models status --json
openclaw gateway restart
openclaw statusopenclaw agent --local --message "Reply with OK only." --thinking off --jsonIf you want users to talk through Telegram, add a Telegram bot token, restart the gateway again, then send a message to the bot.
source ~/.nvm/nvm.sh
nvm use 22
openclaw config set gateway.mode '"local"' --strict-json
openclaw channels add --channel telegram --token "<your-telegram-bot-token>"
openclaw gateway restart
openclaw channels status --probeAdding the bot token is not enough by itself. The gateway must be running, and the user must start the Telegram bot at least once.
Merge the aimesh provider into your existing openclaw.json instead of deleting unrelated providers you still need.