Getting Started with Chat

The chat panel is your always-available interface for working with Agency Hero’s AI agents. Whether you’re deep in a client workspace or browsing your home feed, the chat is right there — ready to answer questions, surface context, take actions, and coordinate complex tasks on your behalf.

This article walks you through the basics: what the chat is, how to navigate it, and how to get the most out of your first conversations.

What is the chat interface?#Copy link

The chat panel is a persistent AI workspace built into the Agency Hero shell. Unlike typical chatbots, it doesn’t just answer questions — it’s grounded in your real workspace data. The agent knows what workspace you’re looking at, can search your meeting history and knowledge base, create tasks, log decisions, and even spawn specialized sub-agents to tackle complex requests in parallel.

Key characteristics:

• Always visible — The chat never unmounts as you navigate. Switch pages, open different workspaces, or dig into a task — the chat stays where it is.
• Context-aware — Every message you send includes information about where you are in the app. The agent automatically loads context relevant to your current workspace, meeting, or task.
• Action-capable — The agent can read and write. It can create tasks, log intelligence items, propose updates, and more — with your explicit approval before any changes are made.
• Multi-threaded — You can run multiple conversations at the same time, each in its own thread tab.

Opening and navigating the chat#Copy link

Work Mode vs. Chat Mode#Copy link

Agency Hero has two layout modes that affect how the chat panel is presented:

Work Mode (default) — The workspace content fills the center of the screen, and the chat panel sits in a column on the right. Thread tabs appear at the top of the chat column so you can quickly switch between active conversations.

Chat Mode — Chat becomes the primary focus. It expands to the center of the screen, and a Threads Rail appears on the left, listing all your conversations grouped by date (Today, Yesterday, Previous 7 Days, etc.). The workspace content moves to a narrower column on the right.

To switch between modes, press Cmd+Shift+C (Mac) or Ctrl+Shift+C (Windows/Linux), or click the Work/Chat toggle in the left navigation rail. Your preference is saved and persists across sessions.

Adjusting chat panel width#Copy link

In Chat Mode, you can resize the chat column using the Size Presets control (the sliders icon in the chat header). Options range from Small (320px) to Maximum (up to 1,200px, filling the full workspace width). Useful presets:

Starting a conversation#Copy link

Sending your first message#Copy link

Type your message in the composer at the bottom of the chat panel and press Enter to send. To insert a line break without sending, use Shift+Enter.

The composer looks like this:

Starting a new thread#Copy link

You can start a fresh conversation at any time:

• In Work Mode, click the [+] button in the tab bar at the top of the chat column.
• In Chat Mode, click [+ New Chat] at the bottom of the Threads Rail, or the Plus icon in the chat header.

Each new thread is a blank slate with no prior conversation history. The agent will still have access to your workspace context — it just won’t have memory of previous chat messages.

Switching between threads#Copy link

• In Work Mode, click any tab in the tab bar to switch. You can have up to 10 tabs open at once. Active streaming is indicated by a spinner on the tab.
• In Chat Mode, click any thread in the Threads Rail. Threads are grouped by date and searchable.

Closing a tab (the × on the tab) doesn’t delete the conversation — it just removes it from your open tabs. You can find it again in the Threads Rail or via search.

What the agent knows about your context#Copy link

One of the most important things to understand about Agency Hero chat is that the agent always knows where you are. Every message you send automatically includes your current location in the app — what workspace you’re in, what page you’re on, and what item you’re looking at.

Here’s how context maps to agent behavior:

This means you rarely need to tell the agent which workspace or meeting you’re asking about — it already knows. Just ask naturally: “What did we decide last week?” or “Summarize the key risks in this deal.”

Context chips#Copy link

When you navigate to a task or meeting and click “Chat about this,” a context chip appears above the composer, showing what’s attached to the conversation:

• 🔒 Task chip — Shows the task title. The agent will use that task as its primary focus.
• 📅 Meeting chip — Shows the meeting title with a “Using meeting context” label.

You can remove a chip by clicking the × on it if you want to broaden the conversation scope.

What the agent can do for you#Copy link

The Agency Hero agent is a general-purpose orchestrator. It handles simple questions directly and, for complex work, delegates to specialized sub-agents behind the scenes.

Reading and answering questions#Copy link

The agent can search across:

• Workspace knowledge — Documents, notes, and extracted insights from your workspace
• Meeting summaries — Synthesized summaries from past meetings
• Meeting transcripts — Detailed utterance-level search within a meeting
• Topics — Structured knowledge topics your team has built up
• Intelligence items — Decisions, risks, open questions, and signals logged in your workspaces
• Tasks — Task lists with filtering by status, assignee, and more
• Upcoming meetings — Your calendar across all workspaces

Taking actions (with your approval)#Copy link

The agent can propose changes that require your explicit approval before they execute. When the agent wants to take an action, you’ll see an approval card inline in the chat:

• Task creation / update — Review the proposed task details, edit if needed, then approve or deny.
• Intelligence items — Log a decision, flag a risk, capture an open question.
• Linear tickets — Propose a ticket with title, description, and priority.

Approval cards show you exactly what will change. You can edit fields before approving. Nothing is written without your confirmation.

Spawning sub-agents for complex work#Copy link

For research-heavy or multi-step requests, the agent can spawn specialized sub-agents. These run in parallel and return results to the main conversation. You’ll see the main agent synthesize everything into a coherent response. This happens automatically — you don’t need to do anything differently.

Workspace vs. cross-workspace context#Copy link

By default, the agent is scoped to the workspace you’re currently in. This keeps responses focused and relevant.

However, you can ask questions that span multiple workspaces and the agent will search across all workspaces you have access to:

“What’s my week looking like across all my projects?”
“Are there any overdue tasks anywhere in my account?”
“Find all the decisions we’ve made about the API architecture.”

When the agent needs to write something (create a task, log an intelligence item), it will ask you which workspace to target if it can’t determine the right one from context.

Thread lifecycle: how conversations are managed#Copy link

Agency Hero manages thread lifecycle automatically so you don’t have to think about it:

You can also manually archive a thread via its context menu in the Threads Rail. Archived threads are never deleted — they’re just tucked away.

Tips for effective use#Copy link

Be specific, but not overly formal. The agent understands natural language well. You don’t need to phrase things like commands — conversational questions work great.

Use the web search toggle for external questions. If you want the agent to look something up beyond your workspace (industry news, a competitor, a technical reference), click the 🌐 globe icon in the composer before sending.

Start fresh threads for new topics. Threads accumulate context over time. If you’re switching to an unrelated topic, a new thread keeps things clean and the agent more focused.

Let context do the work. Navigate to the relevant workspace, meeting, or task before asking a question. The agent will automatically load the right context — you won’t need to explain what you’re talking about.

Review approval cards carefully. Before approving any action, check the details in the card. You can edit fields directly in the card before clicking Approve.

Ask follow-up questions freely. Each conversation maintains its history within the thread. Build on previous answers, ask for clarification, or request a different format — the agent remembers what you’ve discussed.

Keyboard shortcuts#Copy link

What’s next#Copy link

Now that you know the basics, explore what your agents can really do:

• Understanding AI Agents — Learn how the agent architecture works, what skills are, and how the orchestrator delegates to sub-agents.
• Asking Effective Questions — Tips and patterns for getting better answers from your agents.
• How Skill Routing Works — Understand how the agent decides which specialist to delegate to.