IMPLICATOR
HybridClaw is positioning a German-built enterprise agent runtime around controls that the viral Hermes Agent movement still leaves mostly to operators. The open-source package, latest on npm at version 0.13.1, imports OpenClaw and Hermes homes while adding encrypted credentials, approval flows, audit logs and managed cloud options. For companies that watched personal agents turn into permanent workers, the pitch is simple: make the agent boring enough to trust.
Key Takeaways
- HybridClaw is positioning itself as a controlled enterprise runtime around the Hermes and OpenClaw wave.
- Hermes has the adoption graph, with more than 120,000 GitHub stars checked Monday.
- HybridClaw's pitch centers on audit logs, encrypted credentials, migration paths and EU-hosted business workflows.
- The open question is customer proof: local routing and token savings still need production evidence.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Hermes sets the denominator
Hermes Agent is the hype benchmark. GitHub API data checked Monday showed 120,314 stars for the Nous Research project, compared with 83 for HybridClaw. That is not a close race for developer attention. If you run agents inside a company, it is not the only race.
But it is the wrong race for the customers HybridAI wants. Hermes sells a learning loop, memory, skills, messaging gateways and server-side persistence. HybridClaw sells the office version of that idea: a runtime a finance team, support desk or BI group can use without treating every workflow as a weekend experiment.
Co-founder Benedikt Koehler described skills as the real value in these systems during an interview with implicator.ai. His examples were not moonshots. Invoice folders. BI queries. SEO checks. Meeting follow-ups. Annoying work that already sits in email, Excel, Teams and file shares.
Controls come before charisma
Koehler said HybridClaw stays compatible with OpenClaw and Hermes where compatibility matters: skill systems, configuration, provider credentials and memory files. The product diverges at the control layer. Rollback, backup, audit views and permission boundaries matter more when the agent can read company data and run tools.
HybridClaw's own release notes point in that direction. Version 0.13.1, published April 24, added delegated-agent usage logs, token counts, model routing for subagents, encrypted web-search secret propagation and safer WhatsApp gateway shutdown. Dry stuff. Exactly.
The channel list is broad: Teams for office users, email for forwarded work, and the usual chat surfaces from Discord to WhatsApp. The site also says Docker provides full container isolation during local installs, while managed cloud instances run inside secured containers.
Track the Agent Stack
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
The European sell is practical
HybridAI calls the stack EU-hosted and aligned with GDPR and the AI Act. It also points buyers toward VAT classification, business intelligence and compliance checks. The homepage names Dr. Benedikt Koehler, Stephan Noller and Dr. Christian Ross as co-founders. It claims more than 5,000 custom AI agents.
Koehler's interview answer was smaller than the sovereignty label around it. A local model can take the dull passes first. The expensive model comes in later, if the task earns it. His rough figure was 80 percent of tokens staying on the local machine for simple work. Useful, if it survives contact with invoices and support mail.
The log file is the product
HybridClaw is early. Hermes has the crowd, the forks and the release drumbeat. HybridClaw's test is duller and harder: can a controller see the credential, the tool call and the memory write before the agent makes a mess?
That is the German answer here. Check the log. Deny the next click. The agent waits.
Frequently Asked Questions
What is HybridClaw?
HybridClaw is an open-source agent runtime from HybridAIOne that targets enterprise use cases. It supports migration from OpenClaw and Hermes homes, channel integrations, memory, approvals, encrypted credentials and local or managed deployment paths.
How is HybridClaw different from Hermes Agent?
Hermes emphasizes a learning loop, memory, skills and server-side persistence. HybridClaw's pitch is more operational: business workflows, auditability, rollback, credential handling, EU hosting and controls that companies can inspect.
Who is behind HybridAI?
HybridAI names Dr. Benedikt Koehler, Stephan Noller and Dr. Christian Ross as co-founders. Its site says the company focuses on AI coworkers, business intelligence, VAT automation, compliance workflows and EU-hosted infrastructure.
Why does the GitHub star gap matter?
It shows that Hermes has far more developer attention right now. HybridClaw is not winning on community scale yet, so its case depends on enterprise controls, customer deployments and whether buyers value auditability over hype.
What still needs proof?
The strongest open question is production evidence. Koehler said simple tasks can sometimes move about 80 percent of tokens to local execution. That claim needs customer examples before it can be treated as more than an interview claim.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: [email protected]
