❓ Help May OpenClaw user ba dito?

[Draft]

Sa meron nang You do not have permission to view the full content of this post. Log in or register now. experience, ano ang mga useful features niya para sa inyo?
Planning to deploy it on a small VPS.

 

[Cebu]

wow nice thread paps gusto ko din malaman if meron gumagamet. gamet namin yan sa work ko openclaw noon clawdbot hehe

 

[Miskha]

Free lang ba sya lods?

 

[Draft]

Free lang ba sya lods?

Yes free to install.

 

[Miskha]

Yes free to install.

parang automation din ba sya lods?

 

[xLynx]

Present sir. Bakit?

 

[xLynx]

Gamit ko as remote enabler via API endpoit. So part sya ng company ko as personal AI features that we can talk to AI to do something in behalf of us sa mga docs.

 

[Miskha]

Gamit ko as remote enabler via API endpoit. So part sya ng company ko as personal AI features that we can talk to AI to do something in behalf of us sa mga docs.

boss pwede magtanong pano sya iinstall sa mac po? gusto ko sana matry

 

[xLynx]

boss pwede magtanong pano sya iinstall sa mac po? gusto ko sana matry

visit mo lang You do not have permission to view the full content of this post. Log in or register now. may one click install via CLI command. If you are not familiar with cli, better watch tutorial from youtube for easy learning.

 

[Minato NamikaZee]

kung gustomo free lahat boss need mo powerful gpu pra sa local llm tulad ng gemma4, pero if may pera ka nman masgusto mo mas malaksan bibili ka tlagang api, librelangn mn yung openclaw, mahirap lang dinpagnontechy atbagohan ka at may yt tuts boss madali lang yan..

 

[PHC-TheGlock]

I've heard of it before from my friends but di ko pa sya na tatry... can anyone tell me the primary use, and kung saan sya useful?

 

[Graphmaster]

try to xplore this boss draft

 

[alist1986]

Yang openclaw, 24/7 proactive personal assistant (intelligent AI Agent ). Ikaw yung bahala na mag-control ng gagawin nyang tasks. Pwede yang personal up to corporate AI assistant. Endless possibilities, depende sa kung anong i-enable mo or idagdag mong skills nya via clawhub, github repos or create your own. As raw openclaw, you can run it locally with your favorite LLM (online apis or local ai). Kung ayaw nyo ng local, gamit kayo ng custom "claws" ng mga AI providers like Manusclaw, claude cowork, or host it in the cloud.
Bagay lang yan sa mga power users, mahilig sa multi-tasking, mga kalimutin na kailangan sabihan ng schedules, previous data, tulungang magsagot/mag-send ng emails sa reports habang tulog (he he), mag-store ng private data para pwedeng balikan ulit at paalalahanin....Automated repeated tasks ang malimit na gamit nyan though marami pa siyang kayang gawin!
Hindi yan bagay sa'kin. Pero basta alam nyong gumamit ng free MCPs sa inyong app, kaya rin nyang maging openclaw like claude desktop, librechat, opencode, cursor...Basta mga agentic apps na may long-term memory, local execution at auto-loops na nakabaon, ok na yan as alterantive - may UI kapa he he.

 

[alist1986]

Nalimutan kong sabihin ito. Since cintinuously running yan, matakaw malamang sa tokens, so local AI is a must for personal use, Pero doon sa may pambayad just study how to separate llm tasks of the agents by configuring openclaw.json. Below will five you an idea:
You do not have permission to view the full content of this post. Log in or register now.
You do not have permission to view the full content of this post. Log in or register now.
Sample configs:
light cost:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "anthropic/claude-opus-4-5",
        "fallbacks": [
          "ollama-cloud/llama-3.1-70b",
          "ollama/llama-3.1-8b",
          "google/gemini-3-flash"
        ]
      },
      "models": {
        "ollama/llama-3.1-8b": { "alias": "local" },
        "ollama-cloud/llama-3.1-70b": { "alias": "cloud" },
        "anthropic/claude-opus-4-5": { "alias": "opus" }
      },
      "heartbeat": {
        "every": "30m",
        "model": "ollama/phi-3-mini",
        "target": "last"
      },
      "subagents": {
        "model": "ollama-cloud/llama-3.1-70b",
        "maxConcurrent": 3,
        "archiveAfterMinutes": 60
      },
      "imageModel": {
        "primary": "google/gemini-3-flash",
        "fallbacks": ["ollama/llava"]
      },
      "contextTokens": 200000
    }
  }
}

zero cost:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "ollama-cloud/kimi-k2.6:cloud",
        "fallbacks": [
          "ollama-cloud/glm-5.1:cloud",
          "openrouter/openai/gpt-oss-120b:free",
          "ollama/llama-3.1-8b"
        ]
      },

      "models": {
        "ollama/llama-3.1-8b": { "alias": "local" },
        "ollama-cloud/llama-3.3-70b": { "alias": "cloud" },
        "ollama/minimax/minimax-m2.5:free": { "alias": "minimax" },
        "openrouter/nvidia/nemotron-3-super-120b-a12b:free": { "alias": "nemo" }
      },

      "heartbeat": {
        "every": "20m",
        "model": "openrouter/openrouter/free",
        "target": "last"
      },

      "subagents": {
        "model": "groq/llama-3.3-70b-versatile",
        "maxConcurrent": 5
      },

      "imageModel": {
        "primary": "google/google/gemini-3-flash-preview",
        "fallbacks": ["nvidia/google/gemma-4-31b-it"]
      }
    }
  }
}

Meron pang ibang paraan, pero ito yung basic. Bahala kayong mag-edit ng samples to your liking. yung "primary" (main agent) ang pambato mong model, yung "model" is for your custom switching using /model command using thei aliases, yung "heartbeat" ay yung tipong nagpapaalala sa'yo, yung "subagents" are for mini background tasks na kailangan ng main agent, at yung "imagemodel" is for vision/scanning docs. Kung gusto nyo ng TTS, kayo na maglagay nyan.
Dyan ako nahirapan noong una dahil isa lang api ko. Uso naman ngayon yung "compounding apis" kaya madali na. Pero kung may high-end kayong hardware, ollama locally is enough or llama.cpp dahil mabilis yan kaysa python apps.

 

[alist1986]

Ito yung magiging itsura ng .env para humana yang opeclaw.json.

# === CORE SECURITY ===
OPENCLAW_GATEWAY_TOKEN=your_secure_passphrase_here

# === OLLAMA CLOUD (Primary & Fallbacks) ===
# Map to your "ollama-cloud/" JSON prefix
OLLAMA_CLOUD_API_KEY=your_ollama_cloud_key
OLLAMA_CLOUD_BASE_URL=https://your-cloud-endpoint.com

# === OPENROUTER (Heartbeat & Fallbacks) ===
# Handles "openrouter/" prefix
OPENROUTER_API_KEY=sk-or-v1-xxxxxx

# === NVIDIA (Image Model Fallback) ===
# Handles "nvidia/" prefix
NVIDIA_API_KEY=nvapi-xxxxxx

# === GOOGLE GEMINI (Primary Image Model) ===
# Handles "google/" prefix
GOOGLE_API_KEY=AIzaSyxxxxxx

# === GROQ (Subagents) ===
# Handles "groq/" prefix
GROQ_API_KEY=gsk_xxxxxx

# === LOCAL OLLAMA (Local Fallback) ===
# Handles "ollama/" prefix
OLLAMA_BASE_URL=http://localhost:11434

# === SYSTEM SETTINGS ===
OPENCLAW_LOG_LEVEL=info

Samga gustong sumubok ng Ollama + Openclaw, test nyo itong guide na pinagdugrong ko lang at inedit ng kaunti:

Basic installation guide using Ollama with OpenClaw:

A. USING OPENCLAW WITH OLLAMA LOCAL AI
Step 1: Install Ollama and launch OpenClaw with your model of choice
Install Ollama from ollama.com, then use the `ollama launch` command to spin up OpenClaw with a specific local model: using @ollama
https://ollama.com/download
Note: Before this, after Ollama installs completely, go to Settings, click signin with an account and connect Ollama to your system

CHOOSE THE MODEL THAT FITS YOUR HARDWARE:
# 8GB VRAM
ollama launch openclaw --model qwen2.5:7b

# 16GB VRAM
ollama launch openclaw --model llama3.3:70b

# 24GB+ VRAM
ollama launch openclaw --model deepseek-r1:32b

The `ollama launch` command handles everything. It pulls the model if you don’t have it, installs OpenClaw if it isn’t on your system, configures the gateway, and opens the terminal interface.
No manual JSON editing or `openclaw onboard` steps needed.

To check other local ollama models installed, just type "ollama list" or "ollama ls"


B. USING OPENCLAW WITH OLLAMA CLOUD AI:

With ollama+openclaw setup finished, with you signed-in with Ollama online, it will be easy:
1. Go here: https://ollama.com/search
   Search for the cloud models to try, say "https://ollama.com/library/kimi-k2.6"
2. The openclaw command for ollama seen at the bottom of the page to integrate is "ollama launch openclaw --model kimi-k2.6:cloud"
   Openclaw will now use kimi-k2.6:cloud as your model
This is just the basics to use the Ollama AI with Openclaw....
 
Refer to the stan[ATTACH type="full" size="1920x1034"]4173672[/ATTACH]dard openclaw guide for more details on its advanced use: https://docs.openclaw.ai/
--Getting started
--Opening the Control UI
--Run Onboarding
.....

..... Sa terminal nyo na ma-edit yung ibang features nya para magpalit ng models, gumawa ng workspace, magdagdag ng skills, etc.
(Ang sakit sa ulit mag-install ulit at mag-setup he he.)
Try nyo na lang yung ClawX na user-friendly.

Note: ang advantage ng ollama with openclaw setup is bahala yung ollama sa pag-install. Kung may opeclaw ka na, lalaktawan na nyan automatically at supported nya lahat ng OS. Yung manual setup ng openclaw yung mahirap he he. Pag llm lang yung nilagay nyo without setting the tools, chatbot lang yan.

 

[alist1986]

Pahabol ko na lang sa iba:
Mas maganda yang openclaw as self-hosted using docker or thru vps, and safer. This is not for newbies, but you can still try until you fill in the dots. Security ang laging issues sa openclaw!
Kung gusto nyo yung user-friendly na openclaw, look for the ones with a dedicated graphical interface para lantad na lahat ng pipindutin nyo. Marami na rin openclaw clones na mas lightweight, with GUI; hindi ito via TUI, cmd line, or manually configuring config files. You'll need a lot of guides and manuals, as the documentation can't explain fully. Maraming issues din 'yan, so you need to be updated sa kanilang GitHub dev just in case you face a bug. Try to look for full step-by-step guides sa net para medyo malayo yung marating nyo.
For the clones:
ClawX, get it here: You do not have permission to view the full content of this post. Log in or register now.
Search nyo na lang yung iba like NanoClaw, You do not have permission to view the full content of this post. Log in or register now., Nanobot, Zeroclaw.... or look here: You do not have permission to view the full content of this post. Log in or register now.

Doon naman sa katulad ko na papitik-pitik lang ng AI apps sa desktop, I recommend these 2, which can integrate Openclaw in it na may GUI din.
AnythingLLM - via file-based shared storage or a shared local model server
Open-Webui - via You do not have permission to view the full content of this post. Log in or register now. or via openclaw's built-in gateway (Ito yung mas synchronized at bi-directional na setup na may multi-user support pa. Gamit ko ito noong nakaraan pero di ko ma-maintain na 24/7 yung openclaw sa pc ko.)
Yung lumang version ng Jan AI, may integration sa Openclaw pero inalis nila starting from You do not have permission to view the full content of this post. Log in or register now., high memory, and token usage. Claude Code na lang yung natira pero useful pa rin since it can use any AI apis.
Makukuha nyo ang paggamit ng app thru hands-on and patience.
Sa palagay ko, this is enough to know about openclaw for now. Kayo na bahala sa huli.

 

[Draft]

ClawX, get it here: You do not have permission to view the full content of this post. Log in or register now.

Salamat po dito sa tip, mukhang maganda rin pag-praktisan.

 

[Draft]

ClawX, get it here: You do not have permission to view the full content of this post. Log in or register now.

Tested on Ubuntu 26.04, na install naman ng maayos.
Configured using OpenRouter API, pero palaging ganito ang output ng chat :|

 

[alist1986]

Tested on Ubuntu 26.04, na install naman ng maayos.
Configured using OpenRouter API, pero palaging ganito ang output ng chat :|

View attachment 4173973

Buti't gising pa ako. Kalimitan ng ganyang error, nag-exceed ka ng maximum token limit na i-process ng model mo. Pwede rin mangyari yan kung mababa yung setting ng token limit ng model mo sa ClawX. Example, sa Ollama, pag di mo binago, naka-set lang yan sa minimum na 4K (dahil designed sa local AI for chat), so kahit 256k yung model mo, yayariin ka doon sa setting mo sa model kung heavy coding or document processing ang gagawin nya.
Maraming pinang-gagalingan yung error like accumulation ng history na i-resend din kasama yung bago + yung system prompt na kumakain din ng tokens. Yan yung una kong naisip. Kung sumasagot naman pero biglang mag-error. baka makuha sa himas ng token settings, he he. Or choose a model that has a bigger context window. Usually ang natutunan ko sa openclaw (atbp.) ay 64k token limit sa mga +256k models na gamit ko. (Kaya nga sinabi ko noong una, matakaw sa tokens yan, he he.)

Doon sa mga nagbabasa, ito yung sinusunod kong guide sa setting tokens per model para safe, na galing sa AI din:

Token Usage Scenarios

• Standard Chatting: 8K – 16K is plenty. You rarely need more unless you have months of conversation history.
• Heavy Coding: 64K is usually the "sweet spot." This allows you to paste several entire code files and still have room for the AI's response.
• Novel Writing/Long Roleplay: 64K is the minimum. Detailed "World Info" or "Lorebooks" eat context fast; 64K prevents the AI from "forgetting" the beginning of the story.
• Analyzing Documents: 128K+ is better. If you are uploading PDFs or long transcripts, 64K can fill up in just one or two uploads.

Yang 64k ay good for 128k models and above. Pag 128k tokens na, the minimum is +256k models. Just choose the right model for a specific job or task. Distinguish the use of Frontier (Premium), Fast, and Balanced models to manage tokens and costs as well as their supported modals.

Yung context limit na 64k (as example) is the toal memory of the AI, at yung Request limit (maximum response sa single reply ng AI) ay usually nasa 4k (4096 tokens) - na pwede nyong baguhin din as long as the model can handle it.

Sorry sa long reply pakner. Lumipas na kasi antok ko he he.

Kung adventurous ka, try You do not have permission to view the full content of this post. Log in or register now. at yung kayang mga kabaro (Chinese apps). Very clear ang docs at pwedeng i-clone yung openclaw bukod pa sa features nya. Ready to use na siya agad. Bagay na bagay sa'yo. Test ko bukas. Check mo sa github yung releases at baka magustuhan mo.

 

[alist1986]

Ito yung secret tool to save tokens.

Spoiler

Spoiler contents are visible only to Established Members.