The Master Agent

Everything else in the platform runs on a schedule or a queue. The master agent is the exception, it exists for you, responds immediately, and is the natural interface to the whole system. New installs should start here: a ten-minute conversation with the master agent teaches more about your install than any settings page, because it can look at the actual state of your system and answer about it.

Conversations

The Chat tab holds your conversations with the master agent, and you can keep as many named threads as you like, one for operations, one for a project, one for experiments, each with its own history and its own context settings. The conversation picker filters, creates, renames, and deletes threads. Within a thread you get a model picker for choosing what the master agent thinks with, file upload for attachments it should read, microphone input for speaking instead of typing, text-to-speech playback with autoplay if you want answers read aloud, and a cancel button for stopping a response mid-stream.

Every turn you exchange is stored, embedded, and topic-classified by the conversation system, which is what gives threads their long memory. The master agent automatically receives the recent history of the thread it is in, and it can reach back much further on demand: older conversations, topic digests, memory bank entries, and knowledge bases are all a query away.

Context Controls

Each conversation has its own context settings panel, three controls that shape what the master agent sees every turn. The history depth sets how many recent messages ride along. Auto memory query, when enabled, searches the memory bank for entries relevant to your message and includes what it finds, so the agent arrives already informed. Pinned goals and topics keep chosen things permanently in view: pin a topic and its AI-written digest is present in every turn, pin a goal and the agent keeps it in mind across the whole thread. The topic summaries panel beside the chat shows the digests the system has written for the themes in your conversation, and pinning is one click.

What It Can Do

The master agent acts, it does not just answer. Ask it to build something and it queues a task into the coding pipeline; ask it to find something out and it queues the research pipeline. Tell it to pause marketing, or pause everything, and it manages the pause files that control autonomous activity, the same controls documented in pause and safety, driven by natural language. When agents raise flags, items that need your judgment, the master agent routes them to you and records your answers, and an answered question becomes knowledge the system keeps.

It is also the onboarding guide for each area. Setting up the marketing preset, connecting a chatbot platform, or configuring social accounts can all be done as a conversation: the master agent knows what each area needs, reads the relevant per-area guide on demand, checks the current configuration, and tells you exactly what is missing. The Config tab always exists for direct editing, but most owners configure by asking.

How It Knows the System

The master agent's behavior comes from an editable instructions file, its deep knowledge comes from per-area guides it reads when a topic comes up, and its memory of your install comes from the memory bank and the conversation system. This layering keeps it both current and economical: it is not carrying every detail at all times, it knows where every detail lives. Because the instructions are a file you own, documented in instruction files, you can shape how the master agent behaves, what it prioritizes, and how it addresses you.

Like every agent in the platform, the master agent is model-agnostic. The model picker in the chat header switches what it thinks with mid-conversation, and the thread carries on seamlessly, same history, same memory, same abilities. Many owners run a strong model here, the master agent is the highest-leverage seat in the system, and it is the one agent you talk to directly all day.

Working With It Well

Three habits make the master agent dramatically more useful. First, tell it about yourself and your business early, it saves what it learns to the memory bank, and every agent in the system benefits. Second, use named threads to keep long-running subjects together, the topic digests get richer as a thread develops. Third, when something needs doing on a schedule rather than once, ask it to help you set up a proper automated agent for the job, that is what creating agents is for, and the master agent can write the instructions file with you.