Kilo, an AI infrastructure company backed by GitLab co-founder Sid Sijbrandij, has launched KiloClaw, a new managed service that dramatically simplifies the deployment of OpenClaw agents. Historically, getting sophisticated AI agents into production has been a complex and time-consuming process, often involving tedious configuration, dependency management, and system errors. KiloClaw aims to eliminate these hurdles, allowing users to deploy a production-ready OpenClaw agent in under 60 seconds, moving towards what the company terms 'vibe coding' where infrastructure reliability is as crucial as model quality.
The core innovation of KiloClaw lies in its re-engineered sandbox technology. OpenClaw itself is popular for its task-performing capabilities, such as browser control and file management, but deploying it has been the main challenge. KiloClaw addresses this by running on a multi-tenant Virtual Machine (VM) architecture powered by Fly.io. This approach provides a secure, sandboxed environment with managed network security and proxies, preventing common user errors that could expose sensitive API keys or vulnerabilities. This managed service is presented as a more secure and reliable alternative to self-hosted setups.
KiloClaw also tackles the persistent issue of '3 am crashes' often experienced with locally hosted Node.js processes. The service ensures a constant 'always on' state with built-in process monitoring and auto-restart capabilities. This persistence enables several advanced fe atures, referred to as an 'exoskeleton for the mind.' These include scheduled automations, persistent memory through a 'Memory Bank' system that stores context in Markdown files, and cross-platform command execution, allowing agents to be triggered from Slack, Telegram, or the terminal while maintaining a unified state. Furthermore, KiloClaw integrates seamlessly with the Kilo Gateway, offering flexibility to switch between over 500 different AI models from various providers like OpenAI, Google, and Anthropic, without vendor lock-in. To aid users in model selection, Kilo has also released PinchBench, an open-source benchmarking suite designed specifically for agentic workloads, evaluating models on real-world, multi-step tasks.
Fuente Original: https://venturebeat.com/orchestration/kilo-launches-kiloclaw-allowing-anyone-to-deploy-hosted- openclaw-agents-into
Artículos relacionados de LaRebelión:
- AI Agents Peter Steinberger Joins OpenAI
- Quantum Safe Deploy PQC with Cloudflare Now
- AI Agents Autonomy Risks Essential Guardrails Explained
Artículo generado mediante LaRebelionBOT
No hay comentarios:
Publicar un comentario