Agents

An agent is the configured teammate you give a role, tools, context, and ways to start a run. It carries that setup across threads, channels, and triggers so each request can follow your process without rebuilding context from scratch.

What an agent is for

Use an agent when you want Hyperagent to own a recurring responsibility, not just answer one prompt. A RevOps agent might monitor pipeline movement, a Chief of Staff agent might prepare leadership updates, and a Customer Success agent might track account health before renewals.

The request can start in a thread, a Slack channel, an email, a webhook, or a schedule. The point of the agent is that the same teammate shows up each time with the same role, standards, context, and way of working.

What an agent carries

An agent is made of the pieces it carries into each run:

These are not just settings. They define what the agent knows, what it can do, where it can show up, and how much autonomy it has.

The agent is the configured teammate. A thread is where one run happens. A skill is a method the agent can use when the request calls for it. You will see all three together, but they each have a different job.

Where an agent can work

Agents can be reached from many places, but the run still happens in a thread. The thread keeps the prompts, tool calls, decisions, files, and outputs.

Invocations decide where the run starts. Integrations decide where the agent can read context from and deliver the result back. A Slack mention might start the run, Salesforce might provide the data, and the finished update might land back in Slack, email, a Google Sheet, or another connected tool.

Live ModeThreadSlackTelegramScheduledWebhooksEmail

Use Live Mode when you want an agent to stay on, check for changes, and bring something to you before you ask. The agent wakes up on an interval, runs the checklist you give it, and only notifies you when something needs attention.

Live Mode is useful for proactive monitoring: checking for new signals, or keeping an eye on work that changes throughout the day. Instead of waiting for someone to kick off a run, the agent keeps checking in the background.

Setting	What you configure
Interval	How often the agent should wake up and run its checklist.
Checklist	The specific things the agent should monitor, review, or verify.
Heartbeat model	The model used for the recurring background checks.
Run location	Where Live Mode runs should land, such as a dedicated thread created when you enable it.
Communication channel	Where the agent should notify you, such as a Slack DM.
Slack thread replies	When replies in a Slack thread should invoke the agent again.

Start a thread when you want the full Hyperagent workspace in front of you.

From a thread, you can watch the agent generate assets, run live browsing sessions, @mention skills, and keep the output visible as you refine the process.

Threads are best for manual workflows, complex one-off work, and dialing in how an agent should approach a task before you automate it.

Previous slide

Next slide

Deploy agents to Slack so they can work alongside your team where requests already happen. Each agent can have its own identity and role, like @DataAgent for analysis or @CustomerSuccess for account context, summaries, and renewal prep.

Slack can be tightly controlled or more ambient. You decide what wakes the agent up, where it can respond, and where it can listen for context without joining the conversation.

Setting	What you configure
Identity	Use @Hyperagent or create custom Slack identities for specific agents.
Channels	Add the agent to the channels where your team should be able to use it.
Triggers	Choose @mention, topic-based listening, or @mention to start a run with natural follow-up replies in thread.
Agent behavior	Decide whether the agent should respond to other bots.
Listening scope	Choose additional channels the agent can read for context without posting replies.

Deploy an agent to Telegram when work starts from a phone, a quick message, or a group chat. Each agent pairs with a Telegram bot, so you can message it directly for personal workflows or add it to a group where the team can call on it together.

Use a schedule when a workflow should run on a predictable cadence. Set the agent to run every morning, every Monday, the first day of the month, or any other timer that fits the workflow.

Scheduled runs are best for recurring reports, digests, checks, and other repeatable work where the prompt can stand on its own. Each run starts a new thread, so the schedule should include the context and instructions the agent needs to execute without a follow-up question.

Trigger the agent from another system with an HTTP request. Webhooks are best when the run should start from an external service, such as a form submission, product update, or workflow automation.

Give an agent a dedicated email address when you want an inbox to trigger work. Forward a client request, send a report brief, or route a form notification to the agent, and Hyperagent turns that email into a new run.

You control how the email becomes context for the agent: who is allowed to send to the address, whether attachments are included, how the incoming message is formatted in the prompt, and whether the agent can reply back to the sender.

Setting	What you configure
Email name	Create a dedicated address for this agent, with a unique ID appended to keep it private.
Prompt template	Shape how the message is handed to the agent using fields like subject, body, from, and to.
Allowed senders	Limit who can trigger the agent by email address or domain.
Attachments	Decide whether email attachments should be included in the agent prompt.
Replies	Let the agent send a response back to the original sender when the workflow calls for it.

What defines an agent

The reference docs can go field by field. For the mental model, agent settings answer seven practical questions: who is this agent, what can it use, what does it know, how should it produce the output, where can it show up, how much freedom does it have, and how do you review what happened?

Configure the agent once, then let that setup carry across threads, Slack, email, schedules, webhooks, and Live Mode. The cards below follow one RevOps agent example so the settings feel like a real configuration, not a checklist.

The settings tabs are the UI version of this map. Use them when you want to change the agent's role, access, methods, context, autonomy, triggers, or activity history. The reference section can cover the exact fields later; this page is here to explain what each part controls.

How agents improve over time

Agents get more useful as you turn repeated patterns into reusable context.

• Skills capture repeatable methods: a research process, reporting format, writing style, or procedure for using a specific tool.
• Knowledge gives the agent context it can carry forward: your preferences, team details, project background, and the decisions you do not want to repeat.
• Rubrics define what good output looks like, so you can evaluate against a standard instead of relying on vibes.

You can start with just a name and a system prompt. Over time, the agent becomes more useful because the way your team works starts living inside it.

Agent FAQs

An agent is the configured teammate in Hyperagent: its identity, instructions, tools, skills, knowledge, controls, and ways to start a run. Threads are where individual runs happen. Skills are the methods agents can call. As you teach the agent how your team operates, that context carries into the next thread, channel, and trigger.