Open WebUI, the self-hosted AI chat platform with about 158K GitHub stars, published v0.13.1 on GitHub at 11:07 UTC on 2026-11-08. The headline change: on an Open Terminal that offers agents, coding agents such as Claude Code, Codex and Pi now appear in the model picker, where you can chat with them, call them by voice, pick reasoning effort and service tier, and approve their tool requests in the chat. It also introduces Tether, a desktop companion app for macOS, Windows and Linux that ties your own computer to Open WebUI so models and agents can work in your real files and repositories, reachable from any browser including a phone. Tool chats no longer rewrite the system prompt each round, so provider prompt caches can hit; the project says they now cost much less. About 45 minutes later v0.13.2 fixed a UUID error when serving over plain HTTP.
Key Takeaways
- ✓Release: GitHub v0.13.1 shipped at 11:07 UTC on 2026-11-08 and v0.13.2 followed at 11:53 UTC with fixes; the repo has about 158K stars
- ✓Coding agents as models: Claude Code, Codex and Pi appear in the model picker; each conversation uses 1 agent, and agent chats cannot be regenerated, continued or edited and resent
- ✓Tether: a new desktop companion app for 3 platforms (macOS, Windows, Linux); it connects outward so no ports need opening, and sessions expire after 90 days without use by default (SESSION_EXPIRES_IN)
- ✓Cost and speed: tool chats keep a stable prompt prefix so provider caches can hit; large chats open about 2x faster; typing after pasting 16,000 lines drops from 8.6 s to 74 ms per keystroke (about 117x)
- ✓New Auto tool permission (off by default) lets a review model decide whether a tool call needs confirmation; files under 50,000 characters are included in full instead of through retrieval
Key Decision Metrics at a Glance
Not enough VRAM? Compare cloud API and self-hosting costs
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and the Pain Point
Over the past year, coding agents such as Claude Code and Codex have mostly lived in terminals or in each vendor's own IDE or app, while many teams standardise on a self-hosted chat platform. The split means switching to a terminal to make an agent work, and having no easy way to check progress from a phone. Open WebUI is a self-hosted AI platform with about 158K GitHub stars; version 0.13.0 on November 3 added a tabbed workspace plus Open Terminal desktop and browser tabs. Version 0.13.1, published at 11:07 UTC on Sunday, November 8, goes further and brings the coding agents themselves into the chat interface.
Architecture Highlights and Mechanisms
According to the GitHub release notes, on an Open Terminal that offers agents, Claude Code, Codex, Pi and others appear as models in the picker. You can chat with them by text or voice, choose reasoning effort and service tier, and approve each tool request in the chat. Each conversation uses one agent; agent chats cannot be regenerated, continued or edited and resent, and they fall back to the default model if the terminal goes away. The new Tether app is a desktop companion for macOS, Windows and Linux: sign in once and your computer shows up under Your devices in the terminal menu, so models, agents and automations can work in your real files and repositories. It connects outward to Open WebUI, so no ports need opening, and terminal output resumes after a dropped connection. Under the hood, tool chats no longer rewrite citation instructions, memory, the skills list and knowledge into the system prompt each round, so the prefix stays stable and provider prompt caches can hit.
Benchmarks and Measured Results
This is a product update, and no model or agent benchmark scores were published. The performance figures in the release notes are the project's own: large chats open about twice as fast; marking chats read or unread and removing tags is 4 to 6 times faster; typing after pasting 16,000 lines drops from 8.6 seconds to 74 milliseconds per keystroke, about 117x. The claim that tool chats cost much less comes with no percentage, and real savings depend on your provider's cache pricing.
Developer Quick Start
Upgrade to v0.13.2, released at 11:53 UTC, which fixes the "crypto.randomUUID is not a function" error when opening Open WebUI over plain HTTP at a LAN or Docker address. Coding agents need a connected Open Terminal that offers agents; to let agents work on your own computer, install github.com/open-webui/tether and sign in. Sessions expire after 90 days without use by default, adjustable with SESSION_EXPIRES_IN. Admins can enable the new Auto tool permission mode (off by default), but note that it sends the tool, its arguments and your last few messages to the configured scoring API. Authors of custom pipes should check the changed message layout described under Changed in the release notes.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.