Any plans for support on rust-based lightweight claw AI backends

[kcleung]

OpenClaw is great. However, we all know that it is very resource intensive.

Now there are rust-based lightweight versions available, such as ironclaw, zeroclaw, moltis etc

https://www.scriptbyai.com/rust-alternative-openclaw-zeroclaw/

https://github.com/nearai/ironclaw

https://moltis.org/

all of which have a footprint orders of magnitude smaller, thus dramatically slash the running costs / AI token costs.

Any plans to support such lightweight rust-based claw backends?

[pramodavansaber]

Great question @kcleung! We've been watching the Rust-based claw ecosystem closely. ERPClaw's skill architecture is designed to be runtime-agnostic - the modules (SKILL.md, db_query.py, init_db.py) don't depend on OpenClaw-specific APIs, so in theory they can run on any compatible claw backend.

We haven't tested on ironclaw or zeroclaw yet, but it's something we're interested in exploring, especially for edge deployments and cost-sensitive use cases. The main compatibility requirement would be the action routing layer and the skill manifest format.

If you've had success running any claw skills on these runtimes, we'd love to hear about your experience. That would help us prioritize compatibility testing.

[kcleung]

We haven't tested on ironclaw or zeroclaw yet, but it's something we're interested in exploring, especially for edge deployments and cost-sensitive use cases. The main compatibility requirement would be the action routing layer and the skill manifest format.

Well........ the reality is that all real-world use cases are cost-sensitive (i.e. AI-token, CPU, memory and power-consumption sensitive), and cost will be the main barrier of entry for many potential users.

Therefore, out-of-the-box support and integration with a performant rust claw backend is a must for any stable releases of ERPclaw

[shridipavansaber]

@kcleung -- fair pushback, and you are right that cost is not just a nice-to-have consideration. We have logged this as a priority item rather than exploratory.

The current blocker is that ironclaw, zeroclaw, and moltis have not yet stabilized their SKILL.md compatibility layers -- we would rather wait for a stable interface than ship integration that breaks with the next runtime version. That said, we are actively watching all three and have the compatibility work scoped.

In the meantime: if you are already experimenting with ERPClaw skills on any of these runtimes, please share what you find. Real-world data on where compatibility breaks moves this up the roadmap faster than anything else we could do internally.

[kcleung]

Totally agree! Should we document it as an issue and/or as one of the gating TODO items for any stable 1.0 release of ERPClaw?