The Web UI

You sign in with an email and password from the users list, accounts are managed inside the UI itself, and land on the dashboard. The layout is simple: a row of tabs across the top, and a sidebar that stays with you everywhere. This page tours each piece and links to the deeper guides.

Dashboard

The dashboard answers the first question of every visit: what is my system doing? The top shows system state at a glance. Below it, your agents appear as live lanes, every agent you have defined, each with its current status, so a healthy install reads as a calm row of agents ticking on schedule and a problem stands out immediately. The agent list is dynamic: agents you add through the Config tab appear here the moment they exist, and each lane carries the agent's own controls. An activity timeline runs alongside, the curated feed of what has happened, and agent notes from Impulse surface anything the autonomous layer thinks deserves your eyes.

Chat

Chat is your conversation with the master agent: named threads with a picker to filter, create, rename, and delete, a per-thread context settings panel, the topic summaries panel with pinning, file upload, microphone input, a model picker, text-to-speech autoplay, and a cancel button for stopping a response. It is the steering wheel of the system, most things the other tabs do can also be done by asking here.

Files

Files is a complete file browser over the install with an editor built in. Instruction files, configuration, work queues, generated project files, logs, everything is readable and editable without leaving the browser. It earns its keep most on instruction files, editing an agent's role and watching the next tick pick it up, and on reviewing what the coding pipeline produced.

Goals

Goals is the standing agenda: a plain editable list of what matters beyond the day's tasks. Impulse reads it every tick as its priorities, and the master agent keeps pinned goals in view during conversations. Outcome-shaped goals work best, the agents are good at finding steps.

Knowledge

Knowledge is the window into what the system knows: browse and search the memory bank, review the rewards the learning layer has extracted, and manage knowledge bases, create them, import documents, inspect chunks, and prune. A periodic skim here is the fastest read on what your install has been learning.

Config

Config gathers every setting into panels. The Models panel is the operational heart: the task agents editor where you add, edit, enable, disable, and remove agents, live, plus per-pipeline cards for the coding and research pipelines' model and tick settings, and the Impulse controls. API Keys holds provider credentials. The per-area panels, Marketing, Customer Service, Social Media, and Webpages, edit each area's configuration with validation. System covers voice settings and user accounts. Everything the panels write lands in the same plain files documented in configuration, so the UI and hand-editing always agree.

The Sidebar: Feed, Pause, Flags

Three things stay within reach on every screen. The live activity feed streams what agents are doing right now. The global pause button is the platform's biggest switch: one click stops all autonomous activity, and one click resumes it, the full model is in pause and safety. And the flag counter shows how many items await your judgment, clicking it opens the flags panel where each flag carries its type, an unanswered question, a judgment call, or a system observation, its source, and its topic. Answering flags is the most valuable two minutes of an owner's day: every answer becomes knowledge the system keeps.

Working Rhythm

A practiced owner's loop through the UI is short. Dashboard: are the lanes healthy? Sidebar: any flags? Chat: anything to ask or assign? Weekly, add a skim of Knowledge to watch what is being learned and a glance at Goals to keep the agenda honest. The UI is built so that loop takes minutes, the system does the work, the interface keeps you confidently informed about it.