Hermes Agent Tops 22 Trillion Tokens, Overtaking OpenClaw in AI Harness Race
Updated
Updated · The New Stack · Jun 23
Hermes Agent Tops 22 Trillion Tokens, Overtaking OpenClaw in AI Harness Race
1 articles · Updated · The New Stack · Jun 23
Summary
OpenRouter’s late-June app rankings put Hermes Agent first by total tokens, above 22 trillion, after it had already passed OpenClaw in daily token volume on May 10.
Nous Research’s MIT-licensed project is winning on a memory-first design: persistent context across sessions, self-improving skills, and portability across hundreds of models and imported OpenClaw setups.
OpenClaw still leads on distribution, with nearly 380,000 GitHub stars, a broad messaging gateway, and backing from Nvidia’s NemoClaw and Microsoft’s Windows-based Scout enterprise agent.
That split is pushing the contest beyond model choice into the control layer—runtime, governance, identity, and memory—as enterprises weigh managed oversight against self-hosted flexibility.
Security and ownership are becoming the deciding issues, with audits flagging 341 malicious OpenClaw marketplace entries and buyers increasingly focused on who controls agent memory and policy.
After a single hack compromised 770,000 AIs, is the agentic revolution creating an unmanageable security crisis for businesses?
As AI learns your company's secrets, how can you escape the 'behavioral lock-in' that traps you with one vendor forever?
With AI now making B2B purchase decisions, is your business prepared to sell its products directly to machines?
200 Billion Tokens a Day: Hermes Agent’s Rise Over OpenClaw and the Future of Autonomous AI Agents
Overview
In May 2026, Hermes Agent overtook OpenClaw in overall usage, marking a major shift in the autonomous AI agent market. This rise was fueled by Hermes Agent’s rapid development cycle and frequent major releases, each adding significant new features and expanding its capabilities. Notable updates included broader messaging platform support, advanced user interfaces, and powerful new tools for managing tasks and memory. These improvements, combined with a focus on continuous learning and adaptability, helped Hermes Agent attract more users and set a new standard for depth and flexibility in AI agent design.