What OpenClaw costs
per month.

The software is free. The AI model and the server are not.

OpenClaw cost calculator

Model

Per 1M tokens, in and out: OpenRouter list prices, November 5, 2026.

Each message counts as one request.

$1.16

of AI usage a month on DeepSeek V4.1 Flash, at 20 messages a day

Input
39,000 tokens × $0.045 per 1M
0.1755¢
Output
350 tokens × $0.50 per 1M
0.0175¢
Per message
0.1755¢ + 0.0175¢
0.193¢
Per month
0.193¢ × 20 a day × 30 days
$1.16

The median request across Heyron member agents, October 2026. Prompt caching can lower it.

Host it yourself

$6.16 to $21.16

a month for this AI usage and a $5 to $20 server, plus about 1 to 4 hours of upkeep.

Heyron

$49

a month. The computer, its upkeep and $15 of AI usage are included, about 7,800 messages a month on this model.

DeepSeek V4.1 Flash, 20 a day: $1.16 of AI usage a month. Host it yourself: $6.16 to $21.16. Heyron: $49.

How the math works.

About 39,000 tokens in and 350 out per message
The median request across Heyron member agents in October 2026. An agent re-sends its instructions, tools, memory and the conversation each time it answers or takes a step. OpenClaw requests alone were about 38,000 tokens.
List prices, before caching
OpenRouter's prices on November 5, 2026. Most models charge less when the agent re-sends the same context within a few minutes (prompt caching). Higher effort adds thinking tokens, billed at the output price.
A server at $5 to $20 a month
The range published guides give, checked October 24, 2026. Most suggest at least 2 vCPU and 4 GB of RAM, and upkeep takes about 1 to 4 hours a month.
A month is 30 days
And one message is one request, so a job with several steps counts several times. Scheduled jobs and messages from Telegram or WhatsApp count too.

Monthly AI cost by model.

The same math at 10, 30 and 100 messages a day, at list prices. Which model suits which job is its own guide.

AI usage a month by model and messages a day, before caching
ModelPer message10 a day30 a day100 a day
DeepSeek V4.1 Flash0.19¢$0.58$1.74$5.79
Claude Haiku 5.50.41¢$1.22$3.67$12.23
MiniMax M3.1 Flash0.61¢$1.82$5.45$18.18
Gemini 3.5 Flash-Lite1.3¢$3.77$11.32$37.73
GPT-6.1 Terra4.1¢$12.23$36.68$122
Claude Sonnet 5.58.2¢$24.45$73.35$245
Claude Opus 5.516¢$48.90$147$489
GPT-6 Astra41¢$122$367$1,223

What $49 a month covers on Heyron.

With a low-cost model, hosting OpenClaw yourself usually costs less. Heyron is for people who'd rather pay for the setup, the upkeep and the community than run a server.

The computer, kept running
Your agent's own computer, 2 vCPU and 4 GB of RAM, on 24/7. No server bill, and updates are one button.
$15 of AI usage every month
About 7,800 messages on DeepSeek V4.1 Flash or 180 on Claude Sonnet 5.5, at the prices above. Top-ups of $10, $25 or $50 carry over.
Solo Software Studio
A members-only community on Skool for solo founders learning to build and launch apps, with the Build in Public Playbook.

Cancel anytime. New members: full refund within 14 days of your first payment.

Questions about the cost

Is OpenClaw free?

The software is free and open source under the MIT license. Running it costs a computer that stays on and the AI model it calls, unless you run a local model on hardware you already own.

How much does it cost to run OpenClaw per month?

Published guides put a server at about $5 to $20 a month and AI usage at roughly $20 to $100 or more for regular use, so about $25 to $120 in total, plus 1 to 4 hours a month of upkeep. The model moves the total most, which is what the calculator on this page shows. Read the full self-hosting comparison.

Why does an agent use so many tokens?

Every time it answers or takes a step, it sends the model its instructions, its tools, what it remembers about you and the conversation so far. Scheduled jobs and messages from chat apps add requests while you're away, and higher effort adds thinking tokens, billed at the output price.

Does Hermes cost more to run than OpenClaw?

We can't say. Across Heyron member agents in October 2026, a typical OpenClaw request carried about 38,000 tokens, but too few members run Hermes for us to report a typical Hermes request. On either one, the model you pick moves the cost most. Compare OpenClaw and Hermes in detail.