Coming in the Next Two Weeks: Choose OpenClaw, Hermes, or Both — Same Price

Over the next two weeks, every Rapid Claw customer will be able to pick their agent framework from their dashboard: OpenClaw, Hermes Agent, or both running side-by-side on the same managed instance. No price change, no tier gating, no upsell.

What is shipping

Right now, Rapid Claw runs OpenClaw — that is the only option. Starting this week we are rolling out support for Hermes Agent, the open-source agent from Nous Research with a built-in skill learning loop, alongside OpenClaw on the same infrastructure. You will be able to pick one or run both from the same dashboard you already use.

Why we are doing this

Honestly, because there is no one right framework for every job. Hermes has a self-improvement loop that is genuinely useful for clients with repetitive structured workflows. OpenClaw is what we have been running and we will keep building on it. Locking people into one because it is what we happen to sell does not make sense. We would rather let customers pick the tool that fits their actual workflow.

The "both" option is more than just two runtimes installed on the same box. We are wiring them together via ACP — the Agent Communication Protocol — so they actively orchestrate work to each other. Hermes can plan a workflow and hand specific execution steps to OpenClaw via ACP; OpenClaw returns results and Hermes synthesises the output. Or the inverse: OpenClaw runs a deterministic pipeline and calls Hermes via ACP for any classification or reasoning step where a text-only model is the right tool. The agentskills.io format is shared by both frameworks, so skills are portable — a skill built for one can be consumed by the other. And because the two frameworks are on different failure curves, running both in a pipeline with disagreement detection catches errors that a single-framework stack would ship.

I am also aware of the tension here. Both frameworks are open source. Hermes's skill-learning approach might get absorbed into OpenClaw eventually, or vice versa. That is fine — it means customers win either way. We are not trying to build a moat around which runtime we offer. The value we provide is running these things reliably and handling the infrastructure so you do not have to. That does not change based on which framework is inside the container.

If you want to understand how the two frameworks actually differ before deciding, I wrote up the longer technical version: honest tradeoffs from an operator running both. And if you want to see how the ACP orchestration actually works with flow diagrams and three concrete client patterns, the practical setup guide covers that.

Timeline

We are rolling this out instance by instance over the two weeks starting April 7. Target for full rollout is around April 21. We are not doing a big-bang release — each migration gets checked individually before we move to the next batch. If something breaks on one instance, I want to catch it before it affects the others.

What stays the same

Everything that is not the framework choice stays exactly as it is:

One thing worth saying plainly: this is still a small operation. Brandon handles the technical side, I handle the rest. We are not adding staff or changing how we work. This is an infrastructure addition, not a company pivot.

Questions

Go deeper