Four user profiles. Four completely different monthly bills. Here's what OpenClaw actually costs when you add up everything nobody mentions.
Someone asked in the OpenClaw Discord last week: "How much does this actually cost per month?" The first reply said $5. The second said $80. The third linked to the viral Medium post about spending $178 in one week.
All three were honest. All three were describing completely different setups.
The real OpenClaw cost depends on four things: which model you use, where you host it, how many agents you run, and whether you count your own time. Most cost guides cover the first two and ignore the last two. Here's the full picture for four types of users.
The four costs nobody puts in the same place
Cost 1: AI model API fees. This is the variable cost everyone talks about. You pay your model provider (Anthropic, OpenAI, DeepSeek, Google) for every token processed. Range: $0 (Gemini free tier) to $150+/month (Opus on everything, no optimization).
Cost 2: Hosting. Where the agent runs. Range: $0 (your laptop) to $24/month (VPS) to $19/month (BetterClaw Pro). Your laptop costs nothing but only works when it's open.
Cost 3: Your time. Self-hosted setups require 6-8 hours of initial setup and 2-4 hours/month of maintenance (updates, monitoring, security patching, troubleshooting). At $50/hour, that's $100-200/month in labor. Managed platforms eliminate this.
Cost 4: Security risk. Not a dollar amount on an invoice, but a real cost. 500,000+ instances exposed on the public internet. 1,400+ malicious skills on ClawHub. CVE-2026-25253 (CVSS 8.8). CrowdStrike published a full enterprise advisory. The cost of a security incident dwarfs months of hosting fees.
For the complete API cost optimization guide, our optimization post covers how to reduce the model cost component specifically.
Profile 1: The hobbyist ($0-10/month)
You're experimenting with OpenClaw on your own machine. Personal use. Learning. Building for fun.
Model cost: $0-5/month. Gemini 2.5 Flash free tier (1,500 requests/day) covers personal usage entirely. Or DeepSeek V3 with ~10M free tokens/month. If you exceed the free tier, DeepSeek costs under $5/month for moderate usage.
Hosting cost: $0. Your laptop or desktop. The agent runs when your computer is on and stops when it's off.
Time cost: 2-4 hours initial setup. Minimal ongoing maintenance because uptime doesn't matter for personal use.
Total: $0-10/month. This is the "$5/month" answer from Discord. It's real, for this specific profile.
Profile 2: The solopreneur ($25-55/month)
You're running one agent for real work. Customer support on Telegram. Task automation on Slack. It needs to be on 24/7 and respond reliably.
Model cost: $8-15/month. Claude Sonnet as primary ($3/$15 per million tokens), Haiku for heartbeats, DeepSeek as fallback. With session hygiene (/new every 20-25 messages) and model routing, a moderate-usage agent (50 messages/day) stays in this range.
Hosting cost: $6-19/month. Either a Hetzner VPS ($6/month) with self-managed infrastructure, or BetterClaw Pro ($19/month) with zero infrastructure management.
Time cost (self-hosted): 2-4 hours/month of maintenance. At $50/hour, that's $100-200/month in time value. Most solopreneurs don't invoice themselves for this time, so it feels free. It isn't.
Time cost (managed): Zero hours/month on infrastructure. All your time goes to agent configuration and actual use.
Total (self-hosted): $14-21/month visible + $100-200/month time = $114-221/month real cost. Total (managed on BetterClaw): $27-34/month. No hidden time cost.
The solopreneur's OpenClaw cost is either $27-34/month (managed) or $114-221/month (self-hosted, including time). The self-hosted bill looks cheaper. The total cost is higher.
Profile 3: The startup with 3 agents ($75-180/month)
You're running three agents: one for customer support, one for internal operations, one for sales qualification. Different channels. Different models. Different skill sets.
Model cost: $25-45/month total. Three agents, each doing 30-50 messages/day. Sonnet primary, Haiku heartbeats, DeepSeek fallback. Model routing configured per agent based on task complexity.
Hosting cost (self-hosted): $12-24/month for a VPS that handles three agents. One server, three workspaces.
Hosting cost (managed): $57/month for three agents on BetterClaw Pro ($19/agent, up to 25 agents, each billed at $19/month).
Time cost (self-hosted): 4-6 hours/month. Three agents means three sets of skills to maintain, three sets of updates to test, and more complex debugging when something breaks. At $50/hour, $200-300/month.
Time cost (managed): Zero on infrastructure. All time goes to agent configuration.
Total (self-hosted): $37-69/month visible + $200-300/month time = $237-369/month. Total (managed): $82-102/month.
For the managed versus self-hosted comparison with 10 decision scenarios, our comparison covers which approach fits which situation.
Profile 4: The enterprise with 10+ agents ($300-1,500/month)
You're running a fleet. Ten agents across customer support, sales, operations, engineering, and HR. Enterprise security requirements. SAML SSO. Audit logs. Dedicated support.
Model cost: $80-300/month depending on model choice and volume. Enterprise deployments often use Opus for complex tasks alongside Sonnet for routine work. High volume multiplies per-token costs.
Hosting cost (self-hosted): $50-100/month for dedicated servers with redundancy. Plus the salary cost of the DevOps person managing the infrastructure (partial allocation).
Hosting cost (managed): BetterClaw Enterprise starts around $499/month. Includes SAML SSO, audit logs, dedicated CSM, workspace isolation, and granular permissions.
The enterprise math: At 10+ agents, the per-agent managed cost is lower than the fully-loaded self-hosted cost (infrastructure + DevOps time + security overhead + compliance management). The breakeven is around 5-7 agents for most organizations.
For the full enterprise comparison, our comparison hub covers the feature-by-feature breakdown at enterprise scale.
The cost most people discover too late
Here's what nobody tells you about OpenClaw costs.
The Anthropic ban changed everything. On April 4, 2026, Anthropic banned Claude Pro/Max subscriptions from third-party tools including OpenClaw. Users who were paying $20/month for Claude Pro and using it through OpenClaw were forced onto API billing overnight. Some saw their monthly costs triple because API pricing is per-token, not flat-rate.
The viral "I Spent $178 on AI Agents in a Week" Medium post happened to someone running Opus on default settings (no model routing, no session management, no context limits). With optimization, the same usage costs $15-25/month. Without optimization, it costs $150+.
The cost you control: model choice, session hygiene, context limits, heartbeat routing. These are configuration decisions that determine whether you're in the $10/month camp or the $150/month camp. For the six-step cost optimization stack, our guide covers each change in priority order.
The cost you can't control: your time, if you self-host. Updates, monitoring, security, troubleshooting. These hours exist regardless of how well you optimize your API costs.
If you want the model cost optimized automatically (smart context management sends fewer tokens per request) and the infrastructure cost eliminated entirely (no VPS, no Docker, no maintenance hours), give BetterClaw a try. Free tier with 1 agent and BYOK. $19/month per agent for Pro (up to 25 agents, each billed at $19/month). Enterprise from $499/month with SAML SSO and audit logs. The API cost is yours to optimize (BYOK, zero inference markup). The infrastructure cost disappears.
Frequently Asked Questions
How much does OpenClaw cost per month?
OpenClaw itself is free and open source. The real monthly cost includes AI model API fees ($0-150/month depending on model and usage), hosting ($0-24/month), and your time (2-4 hours/month for self-hosted maintenance). A hobbyist using Gemini free tier on a laptop pays $0. A solopreneur on a VPS with Sonnet pays $14-21/month visible plus 2-4 hours of maintenance time. With BetterClaw managed ($19/month Pro), the total is $27-34/month with zero maintenance time.
Why did my OpenClaw costs spike after the Anthropic ban?
On April 4, 2026, Anthropic banned Claude Pro/Max subscriptions from third-party tools including OpenClaw. Users who were paying a flat $20/month for Claude Pro were forced onto per-token API billing. Without cost optimization (model routing, session management, context limits), API costs can reach $150+/month on Sonnet or $400+/month on Opus. With optimization, the same usage costs $15-25/month.
What's the cheapest way to run OpenClaw?
Cheapest by sticker price: your laptop ($0 hosting) with Gemini 2.5 Flash free tier ($0 API). Total: $0/month, but only works when your computer is on. Cheapest for 24/7 operation: Hetzner VPS ($6/month) with DeepSeek ($5/month API). Total: $11/month visible plus 2-4 hours/month maintenance. Cheapest managed: BetterClaw free tier ($0, 1 agent, BYOK with Gemini free tier).
Is BetterClaw cheaper than self-hosting OpenClaw?
For most non-developers, yes. Self-hosted visible costs ($14-21/month for VPS + API) are lower than BetterClaw Pro ($19/month + API). But self-hosting adds 2-4 hours/month of maintenance. At $25+/hour, the total cost of ownership for self-hosting ($114-221/month) exceeds BetterClaw Pro ($27-34/month). For developers who already manage servers, self-hosting is genuinely cheaper on paper.
How much does BetterClaw cost compared to other managed platforms?
BetterClaw Pro: $19/month per agent (up to 25 agents, each billed at $19/month). Free tier available (1 agent, BYOK, no credit card). Enterprise from $499/month. All plans use BYOK with zero inference markup. You pay model providers directly at their rates. BetterClaw charges for the platform only. Smart context management reduces per-request token consumption, which lowers your API bill compared to raw OpenClaw defaults.
