The OpenClaw software is open source (MIT) and free to install. You pay for LLM API usage or subscriptions, optional VPS hosting, and electricity if you run hardware at home.

How much does OpenClaw cost per month?

Many personal setups spend roughly $50–150 per month total: about $30–120 on LLM APIs or subscriptions plus $0–20 for hosting. Light use with local Ollama models can be much lower.

Can I run OpenClaw without API costs?

Yes, if you use local models via Ollama for most tasks. You still need capable hardware; cloud APIs are optional for hard tasks only.

How much does OpenClaw cost?

Monthly breakdown for software, models, and hosting (2026)

Short answer: OpenClaw itself is free (open source). Most people pay for LLM usage and sometimes hosting. A typical active personal assistant lands around $50–150/month all-in; light use or local models can be far less.

For architecture patterns (tiered routing, heartbeats, dedicated machine vs VM), see example setups & model routing.

What you pay for (three buckets)

Cost bucket	Typical range	Notes
OpenClaw software	$0	MIT license; you operate the Gateway
LLM API or subscription	$0–150+/mo	Anthropic, OpenAI, OpenRouter, etc. (providers)
Hosting	$0–20/mo	Home hardware, VPS, or managed
Electricity (home)	~$2–8/mo	Mac Mini / small PC always on (guide)

Example monthly profiles

These are realistic ranges, not quotes—your models and message volume dominate.

Profile	Usage	Rough monthly total
Weekend hobby	Few chats/day, one channel	$15–40
Daily personal assistant	WhatsApp/Telegram, some automation	$50–120
Always-on + cron	Heartbeats, subagents, browser tools	$80–200+
Local-first (Ollama)	Local models + occasional cloud	$5–50 (+ hardware you own)

Subscriptions vs API keys

Subscriptions — Claude Pro/Max, ChatGPT/Codex OAuth (Codex tutorial) cap spend differently than raw APIs.
API keys — Pay per token; predictable for low use, spiky for always-on agents.
OpenRouter / routing — Route cheap models for heartbeats, premium for hard tasks (routing playbook).

How to spend less

Tier models — Cheap model for cron/heartbeats; best model only for user-facing chat.
Use Ollama locally for simple tasks (Ollama setup).
Limit always-on scope — Fewer subagents and browser runs = fewer tokens.
Home host vs VPS — Skip VPS bill if a Mac Mini is already on (FAQ: need a VPS?).
Compare to n8n — Workflow runners bill per execution; agents bill per token (vs n8n).

What managed hosting costs

If you do not want to run a Gateway yourself, managed providers charge their own monthly fee on top of your model usage. Compare options on managed hosting—you trade ops time for cash and less control.

FAQ

Is OpenClaw free? The software is. Models and hosting are not.

Why is my bill spiking? Usually heartbeats or subagents on an expensive default model—see FAQ: high bill and tiered routing in example setups.

Can I use OpenClaw for $0? Possible with local-only Ollama and home hardware you already own; not realistic for heavy cloud models 24/7.