crusoe

The shared toolchain for ComfyUI workflow packages on any NVIDIA GPU host: one installer, pinned packs, declared models from Hugging Face or GitHub, the boot, the driver, and Comfy MCP so an agent can drive the machine. MIT, maintained by SoulCraft.

Let AI agents act on their own, never twice: exactly-once side effects, receipt reconciliation after crashes, and 'done' decided by code. MCP gateway on npm (blackout-mcp). Nemotron on Crusoe.

A self-improving robot maintenance team on Crusoe-hosted Nemotron: rewrites its own playbook, tools and roles, and keeps a change only when a paired unseen-seed gate proves it helps without making robots less safe. Held-out exam 66.5 → 92.8, unsafe actions 21 → 0.

Crusoe Cloud managed inference provider for the Vercel AI SDK

Hugging Face Spaces demos showcasing Crusoe AI managed inference — LLM game battles, model arenas, voice, and agent demos

MLflow deployment plugin for Crusoe AI's managed inference API

Chat with any GitHub repository — a long-context codebase agent powered by Crusoe Managed Inference

Browser voice agent on Crusoe Managed Inference — Pipecat + local Whisper STT + local Kokoro TTS, only one API key needed

An agent that writes its own definition of done, falsifies its own checks, and may not claim success until a model family that contributed nothing re-runs them. Built on Crusoe Managed Inference.

An agent team that trains its own next brain. Self-improving LoRA generations gated by a held-out jury. Recursive Agents Hackathon 2026.

Single-shot Go job that deletes stale manifests from Crusoe Container Registry (CCR)

AYA: software feature recovery, independent evaluation, and measured agent workflow improvement

LangChain integration for Crusoe AI Managed Inference — ChatCrusoe chat model (OpenAI-compatible)

Your own OpenAI-compatible AI gateway to Crusoe Managed Inference — LiteLLM proxy, docker compose, virtual keys, fallbacks