Open WebUI shipped version 0.11.1 on August 25, 2026. It’s a smaller release than July’s interface overhaul, but it adds one capability worth knowing about if you use Open WebUI to drive agentic tool calls: the option to manually approve certain tool calls before the AI executes them. The release also fixes three interface bugs, covering chats with an attached personal terminal and the calendar and events feature.
This is also the first time Open WebUI gets a dedicated write-up on selfhostlab.io. It’s come up before, but only inside articles built around Ollama, the local model runner it’s most commonly deployed alongside as a self-hosted ChatGPT-style front-end.
Human-in-the-loop tool approval
The headline feature in v0.11.1 is human-in-the-loop tool approval. When it’s enabled, Open WebUI can pause before running certain tool calls and wait for you to approve them manually, instead of letting the model execute them automatically the moment it decides to.
That distinction matters more as self-hosted chat interfaces get wired into real actions rather than just generating text: file systems, external APIs, home automation setups, or anything else reachable through function calling. Generating a wrong sentence carries little risk. A tool call that deletes a file, changes a setting, or fires off a request to another service does. Manual approval adds a checkpoint between the model deciding to do something and that action actually happening. For anyone running Open WebUI on a homelab where it can reach other self-hosted services, that checkpoint matters. It’s the kind of control that gets more valuable, not less, as tool use becomes a bigger part of how people actually use these interfaces.
Chats with a personal terminal, fixed
Open WebUI lets you attach a personal terminal to a chat from Settings, letting the model interact with a real shell session. That’s useful for things like running diagnostic commands or checking on a service without leaving the chat window. Before v0.11.1, sending a message in a chat that had a terminal attached this way would fail outright with a “terminal unavailable” error, regardless of whether the terminal itself was actually working. That bug is fixed in this release, so terminal-attached chats should send messages normally again.
Calendar fixes: default date and recurring event times
Two more fixes in v0.11.1 clean up Open WebUI’s calendar and events feature, both tied to timezone handling:
- Default date for new events: creating a new event now starts on today’s date. A timezone calculation bug previously pushed the default forward to tomorrow if you opened the event editor in the evening, or back to yesterday if you opened it early in the morning.
- Recurring event times: recurring events now display at the time you actually configured them for, rather than being recalculated against the server’s timezone and showing up at the wrong hour.
Neither fix changes what the calendar feature does, but both remove a source of quiet, easy-to-miss confusion for anyone scheduling events through it regularly.
Context: July’s interface overhaul
v0.11.1 follows a larger release from a month earlier. Open WebUI v0.11.0, published July 27, 2026, reorganized a chunk of the interface: auto-generated titles for chats and notes got noticeably shorter, archived chats moved out of the main sidebar and into Settings, and chat with a note gained the full capabilities of a regular chat, including model selection, tool access, and file attachments. v0.11.1 doesn’t touch any of that further. It’s strictly a follow-up patch for what shipped afterward.
Updating
The full technical changelog, including the exact commit behind this release, is on the official Open WebUI releases page on GitHub. If you’re running Open WebUI through Docker Compose, which is how most self-hosters deploy it, updating is as simple as pulling the latest image and recreating the container. This release doesn’t carry any extra migration steps beyond what Open WebUI already handles automatically on first launch, and it pairs with whatever version of Ollama you already have running as the model backend.