Hey everyone,
First off, huge credit to the team for rolling out OpenClaw as an Orchestrator. It is a massive step toward making Pickaxe a viable backend for complex, multi-agent workflows.
However, as I dive into architecting decoupled agent loops for a social media copilot application, I’m hitting a hard wall with the current model selector (as seen in image_3704e2.png). Forcing all agentic logic through a single, static provider bottleneck fundamentally limits what OpenClaw can actually do.
If Pickaxe wants to evolve from a prototyping tool into a true COTS-first enterprise backend, we need the ability to point the base URL to a custom routing gateway like Kilo Code or OpenRouter.
Here is why a Custom API Base URL is an architectural necessity:
-
The Worker/Supervisor Cascade: Complex apps require dynamic routing. We need to pass heavy, boilerplate generation to high-volume open-weights models, and reserve frontier models strictly for logic validation. A static model selector breaks this pattern.
-
Decoupled Fallbacks: In production environments, relying on a single vendor’s uptime for an entire agent loop is a massive architectural vulnerability. Gateways handle automatic failovers invisibly.
-
True OpenClaw Alignment: OpenClaw was built on a BYOM (Bring Your Own Model) philosophy to prevent vendor lock-in. Restricting the orchestrator to a gated, internal pricing UI contradicts the exact framework it runs on.
I understand Pickaxe monetizes inference flow natively, but adding a Custom Base URL option for advanced orchestrators would instantly attract developers building real, production-ready SaaS architectures who would otherwise just build natively on Cloud Run.