Claude Just Made OpenClaw 50x More Expensive — What It Means for AI Agents Like Felix Craft

Apr 8, 2026 · Rey Midas

On April 4, Anthropic dropped a bomb: Claude subscriptions no longer cover OpenClaw usage.

If you're running an AI agent on OpenClaw with Claude as the backbone, your costs just changed dramatically. Some developers are reporting potential increases of up to 50x their previous monthly spend.

DHH called it "very customer hostile." The Hacker News thread hit 245+ points. The GitHub Issue got 147+ reactions. And OpenClaw creator Peter Steinberger announced he's joining OpenAI.

Here's what actually happened, what it means for autonomous AI agents, and how to keep your agent profitable. Or skip straight to our free OpenClaw Cost Calculator to see your new costs.

What Changed

Boris Cherny, Anthropic's head of Claude Code, explained that "subscriptions weren't built for the usage patterns of these third-party tools." Agent-style tools like OpenClaw consume tokens far faster than normal chat — running agentic loops that can burn through a month's allocation in hours.

The new reality:

• Before April 4: Your $20 Pro or $200 Max subscription covered OpenClaw usage
• After April 4: OpenClaw usage is billed separately via pay-as-you-go API pricing
• Transition credit: Anthropic is offering a one-time credit equal to your monthly plan cost, redeemable until April 17
• Discount: Up to 30% off if you pre-purchase extra usage bundles

The Real Cost Impact

Let's do the math. A typical OpenClaw agent session might use 50K-200K tokens. If you're running an agent that checks in every 2 hours — like Felix Craft does — that's 12 sessions per day.

For hobbyists who were running agents on a $20/month Pro plan, this is a painful 5-50x increase. For businesses generating revenue with their agents, it's a cost of doing business.

How This Affects Felix Craft ($300K/Month)

Felix Craft — the OpenClaw agent we've been tracking since February — now generates $300,000+ per month in verified revenue (Stripe $100K+ and ETH treasury $94K+). Its operating costs? About $1,500/month.

Even if the pricing change doubles or triples Felix's LLM costs, we're talking about going from $1,500 to maybe $3,000-5,000/month against $300K in revenue. That's still a 60-100x ROI.

But here's the key insight: Felix was already paying API rates. At $300K/month in revenue, Nat Eliason wasn't running Felix on a $20 Pro plan. The pricing change primarily hits smaller operators — the indie builders running agents on subscription plans.

Peter Steinberger Joins OpenAI — What It Signals

In the same week, OpenClaw creator Peter Steinberger announced he's joining OpenAI. He said he "tried to talk sense into Anthropic" but could only delay the pricing change by a week.

His quote was pointed: "Funny how timings match up — first they copy some popular features into their closed harness, then they lock out open source."

What this means for OpenClaw users:

• OpenAI integration will improve. With the creator at OpenAI, expect deeper GPT-4o support in OpenClaw
• Multi-model is now essential. Don't depend on a single LLM provider for your agent
• The open-source ecosystem will strengthen. OpenClaw continues as open source with OpenAI's backing

5 Ways to Keep Your OpenClaw Agent Profitable

1. Switch to Multi-Model Routing

Not every agent task needs Claude Opus. Use cheaper models (Haiku, GPT-4o-mini) for routine checks and reserve powerful models for complex decisions. This alone can cut costs 60-80%.

2. Optimize Your Token Usage

Most agents waste tokens on bloated system prompts and unnecessary context. Trim your SOUL.md. Use structured memory instead of dumping everything into context. Target under 50K tokens per session.

3. Reduce Session Frequency

Does your agent really need to check in every 2 hours? For many use cases, 3-4 strategic sessions per day produce 90% of the value at 25% of the cost.

4. Focus on Revenue Per Token

This is the metric that matters now. Felix Craft generates roughly $200 in revenue per dollar of LLM cost. If your agent is generating $0 in revenue while burning tokens, the pricing change is a wake-up call to focus on monetization first.

5. Use the Right SOUL.md Templates

A well-structured SOUL.md makes your agent more token-efficient. Revenue-focused directives prevent the agent from going on expensive tangents. Clear guardrails = fewer wasted tokens.

The Bigger Picture

This pricing change is a maturation signal. When infrastructure gets more expensive, it forces builders to focus on value creation over token consumption.

The agents that survive this transition will be the ones generating real revenue — not the ones running expensive loops that produce nothing. Felix Craft at $300K/month doesn't care about a 3x cost increase. The hobby agent running on a $20 plan might not survive it.

The lesson: build agents that make money first, then worry about optimizing costs. Every dollar of revenue gives you room to absorb infrastructure changes.

Timeline of Key Events

FAQ

Can I still use Claude with OpenClaw?

Yes. You just need to pay for it separately via API billing, not through your Claude subscription. The functionality hasn't changed — only the pricing model.

Should I switch my agent to GPT-4o?

Consider it, especially with Steinberger at OpenAI. Multi-model routing is the safest bet — use the right model for each task type.

Is OpenClaw still worth using?

Absolutely. OpenClaw has 135K+ GitHub stars, a massive security patch in January, and now OpenAI's backing. The platform is stronger than ever. The LLM cost is just one input — and it's controllable.

How does Felix Craft make $300K/month?

Read our complete case study — it covers the architecture, revenue streams, and exact SOUL.md approach that makes Felix profitable at scale.

← More articles · Explore all tools →