Ollama - Install and Use OpenClaw (Local LLMs)

INDEX

• OpenClaw - Pre-Requisites
• OpenClaw - Install & Configure
• OpenClaw & Ollama - Basic CMDs
• OpenClaw - Models

OpenClaw - Pre-Requisites

To install:

• nodejs
• npm

On the Ubuntu machine itself, run:

sudo apt update
sudo apt install -y nodejs npm

Or:

# 1) Install a supported Node.js version
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt install -y nodejs

# 2) Verify the version is new enough
node -v
npm -v

Fix it with a user-owned npm prefix:

mkdir -p "$HOME/.local"
npm config set prefix "$HOME/.local"
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

npm i -g openclaw@latest

OpenClaw - Install & Configure

QUICK START

ollama launch openclaw

Telegram configuration:

openclaw config set channels.telegram.dmPolicy "allowlist"
openclaw config set channels.telegram.allowFrom '["YOUR_USER_ID"]'

To find your Telegram user ID, use any of these:

• Ask the Telegram bot @userinfobot
• Use a Telegram API tool
• Check your Telegram app details / account info

openclaw config set channels.telegram.dmPolicy "allowlist"
openclaw config set channels.telegram.allowFrom '["123456789"]'

OpenClaw requires a larger context window.
It is recommended to use a context window of at least 64k tokens if using local models.

WEB SEARCH AND FETCH

OpenClaw ships with a bundled Ollama web_search provider that lets local or cloud-backed Ollama setups search the web through the configured Ollama host.

ollama launch openclaw

Ollama web search is enabled automatically when launching OpenClaw through Ollama.

To configure it manually:

openclaw configure --section web

Ollama web search for local models requires ollama signin.

Or just use DuckDuckGo, which is free and does not require sign-in.

CONFIGURE WITHOUT LAUNCHING

To change the model without starting the gateway and TUI:

ollama launch openclaw --config

To use a specific model directly:

ollama launch openclaw --model kimi-k2.5:cloud

If the gateway is already running, it restarts automatically to pick up the new model.

RECOMMENDED MODELS

Cloud models:

• kimi-k2.5:cloud — Multimodal reasoning with subagents
• qwen3.5:cloud — Reasoning, coding, and agentic tool use with vision
• glm-5.1:cloud — Reasoning and code generation
• minimax-m2.7:cloud — Fast, efficient coding and real-world productivity

Local models:

• gemma4 — Reasoning and code generation locally (~16 GB VRAM)
• qwen3.5 — Reasoning, coding, and visual understanding locally (~11 GB VRAM)

More models at ollama.com/search.

NON-INTERACTIVE (HEADLESS) MODE

Run OpenClaw without interaction for use in Docker, CI/CD, or scripts:

ollama launch openclaw --model kimi-k2.5:cloud --yes

The --yes flag auto-pulls the model, skips selectors, and requires --model to be specified.

CONNECT MESSAGING APPS

openclaw configure --section channels

Link WhatsApp, Telegram, Slack, Discord, or iMessage to chat with your local models from anywhere.

STOPPING THE GATEWAY

openclaw gateway stop

OpenClaw & Ollama - Basic CMDs

Base: https://docs.ollama.com/integrations/openclaw

OLLAMA

START

ollama launch openclaw
ollama launch openclaw --config

INSTALL

ollama pull <model_name>

RUN (INSTALL IF NOT PRESENT)

ollama pull <model_name>

OPENCLAW

openclaw onboard                     # Setup Wizard
openclaw configure                   # Specific config
openclaw configure --section web

openclaw gateway start              # Start a managed gateway
openclaw gateway run                # Run the gateway in the foreground

openclaw                            # TUI: Chat in CLI
openclaw dashboard                  # Get HTTP link

STATUS

openclaw status --all
openclaw gateway status --deep     # For details
openclaw memory status --deep

HOOKS

openclaw hooks list               # "ready" means enabled
openclaw hooks enable <name>
openclaw hooks disable <name>

SKILLS

openclaw skills list
openclaw skills enable <name>
openclaw skills disable <name>

openclaw skills search <xxx>      # ClawHub-backed skills
openclaw skills install <xxx>     # ClawHub-backed skills
openclaw skills update <xxx>      # ClawHub-backed skills

FIX

openclaw doctor
openclaw doctor --fix
openclaw security audit --deep

OpenClaw - Models

Model Benchmark Sources

• BenchLM: https://benchlm.ai/
• SWE-Bench: https://www.swebench.com/
• LiveCodeBench: https://livecodebench.github.io/leaderboard.html

VRAM - Rules of Thumb

• Q4_K_M size estimate: ~0.6 GB per billion parameters
• Always leave 20-30% headroom for context (KV cache)
• Example: A 24GB card comfortably runs a ~16-19GB model
• It will NOT run a ~21GB+ model with useful context lengths