HighLevel Managed Agents: What They Do After Level Up

B365 SolutionsFacts checked Oct 31, 20266 min read

Illustration: AI agents moving task cards on a control board

Managed Agents are HighLevel's AI agents that you build by describing the job in plain words. HighLevel's Level Up 2026 recap lists a whole batch of upgrades for them. Agents can now remember people, pause and come back later, start from other apps, split work across subagents, and use a computer, with approval controls. Here's what each one does, in plain terms, plus what it costs.

Key takeaways

  • Managed Agents (formerly "Super Agents") are built from natural-language instructions in Agent Studio, and follow a draft, test and publish cycle.
  • Level Up added memory, scheduled follow-ups, inbound webhook triggers, subagents, custom values and computer use, all named in HighLevel's recap.
  • They're an AI agent product, so they're paid. AI Employee Growth includes 100 runs a month, and Unlimited includes 1,000.

What a Managed Agent is

HighLevel says the Managed Agents builder, "formerly called Super Agents, creates AI agents from conversational instructions instead of manually configured flowcharts." Agents can answer questions, use your Knowledge Bases and take CRM actions such as updating records or triggering workflows. You test each agent in a testing panel before connecting it to live activity.

You'll find them under AI Agents → Agent Studio → Managed Agents. A new unified workspace, with your agent list, chat and logs, and settings in one screen, is available through Labs.

What's new at Level Up 2026

1. Memory and learned procedures

Agents can remember details about individual users (preferences, goals, recent events) and save the steps they used to finish a task so they can reuse them. HighLevel's example: an agent that prepares weekly sales summaries learns your preferred format and the steps involved. You can view, add, edit or delete what it has stored from the Memory tab.

2. Scheduled follow-ups (waits)

An agent can do something, pause, and pick up later with its previous context. Waits range from 5 minutes to 30 days, and steps can be timed after a delay or before or after an event such as an appointment. If an appointment moves, the agent adjusts the reminders that apply; if it's canceled, related follow-ups can be skipped. HighLevel says usage charges apply when the agent resumes and does work, not while it waits.

3. Inbound webhook triggers

Another app can start an agent by sending an HTTP POST to the agent's unique URL. HighLevel lists JSON, form data, XML, plain text and multipart as accepted formats. Token verification and duplicate protection are optional. Unlike workflow webhooks, there's no field mapping: the agent reads the data using its instructions.

4. Custom values and business details

Agent instructions can pull in your sub-account's custom values and business profile details. That lets one agent template work across several sub-accounts, and HighLevel flags missing values before you publish.

5. Subagents

A main agent can split a complex job into focused tasks, hand them to specialist subagents that run in parallel, and review their results before it replies. You can see each subagent's activity and restrict subagents to only the apps and actions their task needs.

6. Computer: browser, files and terminal

With Computer enabled, an agent can browse websites, fill out forms, create and edit files, and run scripts. Two important limits from HighLevel's docs:

  • Approvals: "Ask Before" settings require your OK before actions such as submitting forms, making payments, publishing content or exporting data.
  • Chat and manual runs only. HighLevel says Computer "is not supported for trigger-based executions."

7. Import from n8n (beta)

You can turn an n8n workflow export into a Managed Agent. See our n8n import guide.

What Managed Agents cost

HighLevel's AI pricing puts Managed Agents in the paid AI Agents group, even for agency users. The AI Employee plans (per sub-account) include:

Plan Managed Agents included
AI Employee Starter ($20/month, Founder-plan agencies only) Not included; billed pay-per-use
AI Employee Growth ($50/month) 100 runs per month
AI Employee Unlimited ($97/month) 1,000 runs per month

When an allowance runs out, HighLevel pauses extra usage by default. Agencies can turn on Extra Usage in AI Suite to continue at pay-per-use rates.

Safety controls worth setting up first

  • Turn on Ask Before for every sensitive action when you enable Computer.
  • Give subagents only the permissions their task needs.
  • Use the testing panel with a test contact before you publish.
  • Review stored facts in the Memory tab, especially for agents that talk to many users.

A starting use case

HighLevel's own example for scheduled follow-ups is a sales-call agent:

  1. One hour before the call, prepare a company brief.
  2. Thirty minutes after, create a follow-up task.
  3. Three days later, create a manager review task if the deal hasn't moved.

That's a contained, internal job. It touches your CRM, not your customers, which makes it a sensible first agent to test.

Frequently asked questions

What are HighLevel Managed Agents?

They're AI agents you build in Agent Studio by describing the job in plain language. HighLevel says they were formerly called Super Agents and can answer questions, use Knowledge Bases and take CRM actions.

Are Managed Agents free?

No. HighLevel lists Managed Agents as a paid AI agent product. AI Employee Growth includes 100 runs a month and Unlimited includes 1,000. Otherwise, usage is billed pay-per-use.

Can Managed Agents browse websites?

Yes, with the Computer capability enabled. HighLevel says Computer currently works for chat and manual runs only, not trigger-based runs, and Ask Before settings can require approval for sensitive actions.

How do inbound webhook triggers work?

An external app sends an HTTP POST request to the agent's unique URL, and the agent receives the data as its first input. Token verification is optional.

Do Managed Agents remember past conversations?

Yes. HighLevel says Memory lets agents keep details about individual users and save learned procedures, and you can view, edit or delete stored facts in the Memory tab.

How long can a Managed Agent wait before following up?

HighLevel says scheduled waits can range from 5 minutes to 30 days.