Skip to main content

Documentation Index

Fetch the complete documentation index at: https://hyperfx.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

Agents on Hyper are virtual teammates with built-in guardrails and domain expertise. They connect to your tools, follow rules, have their own computer, remember context, and ship work in chat — or run on their own as scheduled tasks, webhooks, or in response to app actions. This page walks through everything you can do with them.

Your default Hyper agent

Every workspace gets a default agent named Hyper. It’s there from day one — no setup. Open a New Chat and you’re talking to it. From the Agents page you can manage every agent in your workspace; from the agent selector at the top of any chat you can switch which one you’re talking to. You can spin up new agents at any time — by hand from the Agents page, by asking Hyper to build one for you, or by installing a template from Discover.

Chatting with your agent

When you talk to an agent in chat, three things show up in the message composer:
  • Agent selector — switch which agent you’re talking to
  • Tool selector (the +) — turn on or off any tool the agent has available for this conversation
  • Suggested prompts — pre-written examples that pre-populate the input
As your agent works, it expands its actions inline. Click any tool call to see exactly what happened — what the agent read, the parameters it sent, the response it got. If a skill was used, you can click into it and see the SOP the agent followed. This is your real-time observability of the agent’s work, before you even leave the chat.

Tool permissions, mid-chat

Click the + in chat and select a tool (e.g. Gmail) to see every action the agent has access to and the permission setting on each one. From here you can:
  • Disable an action entirely (the agent can’t perform it from this chat)
  • Require approval before an action runs (the agent has to ask you first)
The fastest way to learn what an integration can do is just to ask: “Hey Hyper, can you explain to me what you’re able to do with the TikTok and Instagram integrations? Keep it concise.” Hyper will list every action and permission setting it has.

System prompt & model

Each agent has a system prompt — its core set of instructions. Think of this as the agent’s job description. You can edit it any time. You also choose the model the agent runs on. Two recommendations:
  • Haiku 4.5 — fast and cheap, great for simple tasks
  • Sonnet — best for building new agents, creating tasks, and complex multi-step work
You can swap models at any time. If an agent is doing serious work, default to Sonnet.

Memory

Agents on Hyper remember context across conversations and tasks. There are two memory layers:
  • User profile memory — what Hyper knows about you (your role, preferences, brand voice, recurring people in your life). Visible from the agent settings panel — you have full access to read and edit everything stored.
  • Agent memory — what this specific agent has learned over time across the work it’s done. Also fully visible and editable from the agent settings panel.
Both improve with use. If you ever want an agent to remember something specifically — just tell it. “Remember that our Q4 launch date is December 5.” It’ll save it.

The resources tab

Open any agent’s settings panel and you’ll see a Resources tab. This is the agent’s working context — everything that’s “in arm’s reach”:
  • Tasks assigned to this agent
  • Skills attached
  • Files attached as context
  • Tables the agent operates against
  • Knowledge bases the agent can read from
  • Browser contexts the agent can use
Anything attached here is right there in the agent’s wheelhouse. Hyper has a native file system — agents always have access to the whole thing — but pinning files or skills to the resources tab puts them front-and-center for that specific agent.

Tasks

A task is a job an agent runs on its own — without you sitting in chat. Tasks have three trigger types:
TriggerWhen it runs
ScheduledOn a cadence you set (daily at 2pm PST, weekly on Mondays at 4pm, etc.)
WebhookWhen something hits a webhook URL (form submission, Zapier, custom code)
App actionWhen an event happens in a connected app (e.g. every new email in Gmail, every new Stripe payment, every new Slack mention)

Creating a task is just sending a message

The fastest way to create a task is to tell your agent. Example:
Hey, can you create a new task to monitor Reddit every day at 2pm PST for
mentions about Hyper, AI marketing agents, and competitor moves in Google
Ads / Meta Ads. Assign it to yourself.
The agent will set up the task end-to-end — instructions, schedule, trigger — and confirm. Open the Resources tab and the new task is right there, ready to run.

Editing tasks

From the Resources tab, click the three dots next to any task to edit, activate / pause, or delete it. Click into a task to see its instructions, schedule, and trigger configuration in detail. You can change the trigger type any time — flip a daily report into a webhook-driven one, or have it run only when a specific event happens in Gmail.

Skills

Skills are reusable capabilities that empower agents to operate like experts in a domain. Think of them as SOPs or framework instructions — “here’s how to run a competitive ad scan,” “here’s how to generate on-brand ad creative,” “here’s how to check if our brand shows up in AI search.” Every Hyper agent automatically searches the official skill library before doing complex work — they pick the right skill for the job without you having to wire it up. Official skills carry an Official badge and are maintained by us. They’re high quality and constantly improving.

Attaching skills to a specific agent

You can also pin specific skills to an agent’s Resources tab. Reasons to do this:
  • It’s a custom skill you wrote and you want this agent to always use it
  • It’s a niche skill that wouldn’t be picked up automatically but matters for this agent
  • You want belt-and-suspenders — guarantee the skill is in arm’s reach
To attach: hit the + button in the chat composer, choose Skills, then browse and select from the skill library. Or do it directly from the agent’s Resources tab.

Creating your own skills

The best way to write a skill is to ask Hyper:
Hey, can you create a skill for generating on-brand ad creatives? Here's
the process I'd like you to follow: [...]. Let's iterate on it and build
it out together.
Hyper will draft, refine with you, and save the skill. You can also import skills from GitHub — paste a URL and Hyper will pull it in. For a deeper look at the Skills system (browsing, importing, the Files → Skills sub-tab), see Files.

Channels

Bring your agent into the tools your team already uses. The biggest one is Slack. To add an agent to Slack:
1

Open the agent's Channels section

From the agent settings panel, scroll to Channels and click the + icon.
2

Name and configure the channel

Give it a name (e.g. “Hyper Slack”), set the type to Slack, and pick the workspace.
3

Authorize Slack

Sign in and grant access. The agent is now reachable in Slack — DM it directly or @-mention it in any channel it’s added to.
Slack is the most-used channel today; more channel types are rolling out.

Inbox & history

You have full observability over every agent — whether you’re chatting with it directly or it’s running tasks in the background.
  • Agent inbox — every chat and action history for this specific agent. A clock icon next to an entry means it was a scheduled task; no clock means it was a one-off chat.
  • Sidebar history — chat and action history across every agent in your workspace. Click “More” to expand and filter.
Both views support the same filters:
  • Filter by agent (just this agent, or all of them)
  • Filter by tasks only (skip the one-off chats)
  • Filter by channel — for example, see only the conversations that happened in Slack
Click into any task run and you see the full transcript of everything the agent did — same view as you’d see if you watched the agent work in chat. This is your audit trail.

Discover — templates and example prompts

The Discover page is the gallery of agent templates and pre-built workflows. Browse, install, customize.
  • Click any template → Install template → the agent now exists in your workspace, with its tasks, skills, and resources already configured.
  • Send the prompt the template ships with and Hyper helps you customize it for your context.
  • Beneath the chat input, you’ll find example use cases — click one and it pre-populates the chat with the exact prompt you’d send.
For more, see Discover.

Going further