Open WebUI 0.10 ships computer agent mode
Open WebUI v0.10.0 adds a computer agent, team folder sharing, streaming reasoning display, a full memory overhaul, and a webhook event plugin system.
Open WebUI released version 0.10.0 on June 29 with a substantial set of new capabilities.
Computer agent support is the headline change: the Open WebUI Computer lets any chat session run full agent sessions on the local machine with file, terminal, git, and web access, connected through an OpenAI-compatible gateway. This mirrors managed computer-use products from Anthropic and OpenAI, but runs entirely on-premises.
Team folder sharing adds read and write access controls for shared workspaces, addressing the single largest barrier to Open WebUI in team settings. Automatic context compaction handles long conversations without manual window management.
The memory system was rebuilt from scratch with two distinct types: long-lived personal memories and per-conversation context, each managed separately. Administrators can now toggle memory context and configure authentication (LDAP, OAuth/OIDC) from the admin panel, removing the need to edit config files.
An Event function plugin primitive hooks into system-wide events such as signups, uploads, and config changes, enabling Python automation at the platform level. Hybrid search on large knowledge bases now runs in-database, which the release notes describe as dramatically faster.
What this means
The computer agent now makes Open WebUI a viable self-hosted competitor to Claude Desktop for teams that cannot send data to third-party APIs. The event system and team sharing are signals that 0.10 targets organizations, not just individual developers.
For background on securing an agent that can touch your local machine, the guide on running coding agents remotely covers isolation and review practices that apply equally to local computer agents.