openscreen Releases

ZeroClaw v0.8.5 is a security and connectivity release introducing ZeroRelay and ZeroRouter, expanding live chat capabilities, and tightening plugin boundaries.

Sep 5, 2026

This release candidate provides an Intel XPU (SYCL) wheel for LMCache, built against pinned versions of `torch-xpu` and the oneAPI toolchain, designed to integrate seamlessly with upstream Intel vLLM XPU containers.

Sep 1, 2026

This release candidate offers a ROCm wheel for LMCache v0.5.5rc3, specifically configured for AMD ROCm 7.2.4 and PyTorch 2.10.0, supporting Instinct gfx942 and gfx950 GPUs with a specific container environment.

Sep 1, 2026

This release candidate provides a standard ROCm 7.2 wheel for LMCache v0.5.5rc3, optimized for AMD Instinct gfx942 and gfx950 GPUs and ABI-matched to the upstream `vllm/vllm-openai-rocm` image.

Sep 1, 2026

This release fixes a startup failure issue by preventing the application from crashing when the Remote Desktop Services (RDS) certificate download encounters a lack of internet access.

Aug 26, 2026

This update introduces testing infrastructure for the App Builder and adds a Personal Access Token (PAT) feature to enhance security and access management.

Aug 26, 2026

A major performance optimization release featuring 584 commits, focusing on Kimi-K3 and DeepSeek V4. It includes extensive ROCm support, tiered KV cache offloading, and the maturation of the Rust frontend with gRPC capabilities.

Aug 26, 2026

This release focuses on stability and UI improvements, including error boundaries for better crash handling, UI refinements for charts and menus, and various bug fixes for the App Builder and database interactions.

Aug 25, 2026

This update introduces a schema locking feature for table columns and resolves several issues including Cloud plan app history restrictions, debounce logic, and mobile layout bugs.

Aug 25, 2026

A major update introducing six new AI vendors, expanded model support, a multi-vendor transcription framework, and enhanced code sandboxing and reasoning controls.

Aug 21, 2026

Minimal release notes: "This is a patch release on top of v0.27.0. - Support quantized DSpark Markov heads (#50424)"

Aug 11, 2026

A significant release introducing Kimi K3 support across the full stack and expanding Model Runner V2 to non-generative workloads like encoders and pooling, alongside a PyTorch 2.13.0 upgrade.

Aug 10, 2026

This release fixes the Hugging Face paraformer-en alias, adds LoRA fine-tuning support for Qwen3, and introduces stable JSONL output for various binaries.

Aug 4, 2026

ZeroClaw v0.8.4 is a maintenance and hardening release expanding memory and SOP control planes while improving provider reliability and strengthening sandbox boundaries.

Aug 2, 2026

This release introduces a text-to-speech toolkit for Smallest AI, OpenSearch vector database support, and performance optimizations for AgentOS metrics.

Jul 30, 2026

Launches FunASR deployment product hub and improves MCP discovery. Includes runtime downloads for various platforms.

Jul 27, 2026

Introduces AgentOS Tools for platform operations and enhances tracing capabilities. Updates Moonshot model defaults and adds thinking mode support.

Jul 27, 2026

Focuses on the Inkling model family, DeepSeek-V4 performance optimization across vendors, and the maturity of KV offloading and tiered storage. Includes Transformers 5.13.0 upgrade.

Jul 27, 2026

Fixes serialization issues for nested executors and adds a new `TrustedRouter` model class. Revamps entity memory for the second brain.

Jul 26, 2026

Updates the FunASR MCP Server runtime and improves documentation for transcribe_audio usage.

Jul 26, 2026