Automation and integrations
Open WebUI Computer keeps working when the browser tab is closed. You can message it from Telegram or Slack, run tasks on a schedule, trigger tasks from other systems with a webhook, expose workspaces as OpenAI-compatible models, and give the agent extra tools via MCP or OpenAPI servers.
Everything in this section runs without you watching: nobody is there to click Allow, so bot messages, scheduled runs, webhook triggers, and gateway requests execute file edits, shell commands, and tool calls without a per-action gate. Point them at a workspace you're comfortable letting an agent act in, and write prompts that say exactly what to do. For interactive approvals, use the web UI; see approvals and plan mode.
| I want to... | Go to |
|---|---|
| Chat with my computer from Telegram, Discord, Slack, WhatsApp, or Signal | Message your computer |
| Run a prompt every morning, hour, or Monday | Scheduled tasks |
| Trigger a task from CI, a cron job, or another service | Scheduled tasks → webhook trigger |
| Get pinged when a chat or scheduled run finishes or fails | Notifications and webhooks |
| Use a Computer workspace as a model in Open WebUI | Use a workspace from Open WebUI |
| Give the agent tools from MCP or OpenAPI servers | MCP and OpenAPI tool servers |
All of these produce real chats in the workspace sidebar, so whatever runs while you're away is there to review when you get back.