Now onboarding for early access
Compose, deploy, and operate AI inference stacks.
Mantler is a workbench for building and running LLM stacks — what we call mantles — on hardware you control. Register machines, compose runtime and model, and expose as an OpenAI-compatible endpoint.
Machines
All your machines, one place.
Connect local workstations, cloud VMs, or any hardware running the mantlerd agent. Monitor GPU state, installed runtimes, and deployed stacks — local and remote — from one dashboard.
Compose and deploy
Build stacks that actually run.
Assemble runtime, model, and optional layers in the Forge. Compatibility resolves from curated rules and community data — no guessing. Deploy to any registered machine in one action.
Endpoints
Any stack, any client.
Expose any stack as an OpenAI-compatible API with org-scoped keys and usage logging. Use it privately in your IDE, share it with teammates, or open it up to paying users.
Earn from your hardware
Rent out what you've built.
Any endpoint you publish can be made available on the Mantler marketplace. Set your own price, control who gets access, and earn from inference capacity that would otherwise sit idle.
Security
API keys are org-scoped with optional per-stack restrictions. Machine auth uses rotating bearer tokens. Multi-stage inference runs with encrypted payloads on your hardware. How Mantler handles security →