# Open WebUI v0.13.1 introduces Tether desktop app and coding agent support

> **Key takeaway:** Open WebUI expands its local AI capabilities with a new cross-platform desktop companion and specialized coding agents.

**Published:** 2026-11-08T11:07:53+00:00 
**Source:** Open WebUI Releases 
**Category:** ai-local-edge 
**Canonical URL:** https://fosswire.org/news/open-webui-v0131-introduces-tether-desktop-app-and-coding-ag.html 

## Summary
Open WebUI has released version 0.13.1, introducing significant architectural improvements and new ways to interact with local and remote AI models. A major addition is Tether, a new desktop companion application for macOS, Windows, and Linux that bridges a user's local machine with the Open WebUI interface. This allows models and coding agents to interact directly with local files, repositories, and tools without requiring open network ports.

The update also integrates specialized coding agents like Claude Code, Codex, and Pi into the model picker for use within an Open Terminal. Beyond new features, the release focuses heavily on performance optimization, including faster chat loading, improved RAG (Retrieval-Augmented Generation) handling for small files, and reduced costs for tool-based conversations through better prompt caching utilization.

## Technical improvements in Open WebUI v0.13.1
- **Tether Desktop Integration:** A new companion app for Linux, macOS, and Windows that establishes a connection to Open WebUI, enabling models to access local files and tools via a secure session that does not require opening incoming ports.
- **Agentic Workflow Enhancements:** Added support for coding agents in the model picker, allowing users to select reasoning effort and service tiers, while implementing a fallback mechanism to default models if a terminal connection is lost.
- **Prompt Cache Optimization:** Reduced tool-use costs by preventing the constant rewriting of citation instructions, memory recall, and attached knowledge into the system prompt, allowing providers to utilize prompt caching more effectively.
- **Automated Tool Governance:** Introduced an 'Auto' permission mode for administrators, which utilizes a configured scoring API and a review model to automatically execute tool calls that do not require human confirmation.
- **RAG and Performance Tuning:** Implemented a new threshold (default 50,000 characters) where small files are included as full text rather than through retrieval, and optimized large chat loading and sidebar responsiveness through new database indexes.

## Why it matters
The introduction of Tether significantly lowers the barrier for developers to use local LLMs for complex repository-level tasks without compromising network security. These updates move Open WebUI closer to a full-featured local AI workstation, providing the performance and automation tools necessary for professional software engineering workflows.