GoHighLevel Agent Studio tutorial: build a Managed Agent start to finish

By / Published Nov 9, 2026 / Updated Nov 9, 2026

Heads up: the trial and plan buttons on this page use my affiliate links. If you sign up through one I get a commission. It doesn't cost you more.

To build an agent in GoHighLevel Agent Studio, open the client's sub-account, go to AI Agents > Agent Studio > Managed Agents, click + Create Agent and describe the job in plain English: what starts it, which knowledge base it reads, and which CRM actions it's allowed to take. The builder drafts the instructions, triggers and tools, you test it in the test panel, then switch it from Draft to Published, and every run is billed per run (100 runs a month are included on the $50 AI Employee Growth plan, as of Nov 9, 2026).

Quick heads up before we start: Agent Studio isn't AI Studio. AI Studio builds websites (that's AI Studio tutorial), Agent Studio builds agents that work inside the CRM. And since the flow-based canvas stopped taking new agents, "Agent Studio tutorial" now really means Managed Agents, which used to be called Super Agents. Below is one agent built start to finish for a local business, how to hand work to it from a workflow, what each run costs, what's still beta or Labs, and how to copy it into other client accounts.

Start your 30-day free trial →

30 days free instead of HighLevel's usual 14, plus a 1:1 setup call with my team, weekly live calls and my agency resources from day one. See everything you get →

The video version

Short version

  • 01Agent Studio is the home for HighLevel's AI agents. New agents are Managed Agents (formerly Super Agents). The old flow canvas only edits agents you already have.
  • 02AI Studio is a different product. It builds websites and apps from a prompt.
  • 03Build: AI Agents > Agent Studio > Managed Agents > + Create Agent, describe the job, check the triggers and tools it drafted, test, publish.
  • 04Cost: one form submission, tag change or chat session is one run. Growth ($50 a month per sub-account) includes 100 runs, Unlimited ($97) includes 1,000, and anything else is pay per use by tokens. 30 test runs a month are free (prices as of Nov 9, 2026).
  • 05A workflow can hand a task to a published agent with the Invoke Managed Agent action, then use the agent's answer in the next steps.
  • 06Labs or beta as of Nov 9, 2026: the Unified Workspace, Memory and the n8n importer. Computer (browsing, files, Python) works in chat and manual runs, not in trigger runs.

Build a lead triage agent in Agent Studio, step by step

  1. 01

    Check access and the plan

    Work in the client's sub-account with a user that can manage Managed Agents. Decide who pays: without an AI Employee plan every run is pay per use from the wallet, Growth includes 100 runs a month and Unlimited 1,000 (cost details below).

  2. 02

    Pick one job

    We're building a lead triage agent for a local plumbing company. When the website quote form comes in, it reads the answers, checks the job and service area against the knowledge base, writes a note on the contact, tags the lead hot, warm or out of area, and creates a call task for the owner on hot leads. It doesn't text the lead. That stays in a workflow, so A2P rules and Stop on Response still apply.

  3. 03

    Build the knowledge base

    AI Agents > Knowledge Base, create one called Services and Service Area. Add the services page with a web crawl, plus a short table or FAQ with the zip codes they cover and what counts as an emergency (burst pipe, no hot water, sewage backup). Use the Knowledge Base Retrieval Tester to check it finds the zip codes.

  4. 04

    Describe the agent

    AI Agents > Agent Studio > Managed Agents > + Create Agent. Something like: "Create a Lead Triage agent for ABC Plumbing. When the Website Quote Request form is submitted, read the answers and use the Services and Service Area knowledge base. Add a note to the contact with the job, urgency, address and anything missing. Add the tag lead-hot for emergencies, lead-warm for quotes, and lead-out-of-area if we don't cover their zip code. For lead-hot, create a task for the owner due in 15 minutes. Never message the contact and don't change any other fields." That last line is the boundary HighLevel recommends writing in.

  5. 05

    Check what the builder drafted

    Read the instructions, the Form Submitted trigger, the tools and the knowledge base in the configuration panel. If it picked the wrong form, it asks which one when names clash. Fix things with targeted edits like "Make the note shorter without changing the triggers or knowledge base", or edit the field directly.

  6. 06

    Test it on your own contact

    Run the test with a contact that's you, not a real lead, because tests can take real actions. Try an emergency in the service area, a quote, an out-of-area zip and a form with half the answers missing. Open each run's log to see the steps, tools and tokens it used. Test runs come out of the 30 free a month.

  7. 07

    Publish

    Switch the agent from Draft to Published. Only now does the form trigger fire for real leads. If something goes wrong later, Unpublish stops the triggers and clears any scheduled runs but keeps the whole setup, so you can fix it and publish again.

  8. 08

    Add the follow-up workflow

    Automation > Workflows, trigger Contact Tag added: lead-hot. Send the owner an internal notification by SMS and send the lead a text right away. For lead-warm, a slower nurture. The tags the agent writes are what start these, so the agent decides and the workflow talks.

  9. 09

    Watch the Activity tab

    For the first week open the agent's Activity tab every day. Each run shows its status, what triggered it and the full execution log, and you can thumbs-up or thumbs-down replies. Tighten the instructions where it guessed wrong, test, and publish again.

Agent Studio, Managed Agents, flow agents and AI Studio: what's what

The names are a mess, so here's the plain version, from HighLevel's own docs as of Nov 9, 2026.

  • Agent Studio: the area under AI Agents where agents live. It has two tabs, Managed Agents and Flow Agents.
  • Managed Agents: the current builder. You describe the job in words and it builds the agent. This is what HighLevel calls Super Agents in older videos.
  • Flow Agents: the original drag-and-drop canvas with LLM, API, MCP, Knowledge Base and Web Search nodes. HighLevel stopped new flow agents. The ones you built keep running, and you can still edit them.
  • AI Studio: not an agent tool at all. It builds websites, landing pages and small apps from a prompt. HighLevel's AI tools doc says it straight: "AI Studio and Agent Studio are separate products."
  • Conversation AI and Voice AI: the agents that talk to customers over text and phone. They sit next to Agent Studio under AI Agents, and they're billed on their own.

When to use a Managed Agent (and when not to)

Use a Managed Agent for work that happens behind the scenes when something happens in the CRM: a form comes in, a tag changes, an opportunity is won, or it's 8am on Monday. It reads, decides, and updates the CRM.

If the job is chatting with leads over SMS, DMs or web chat and booking them, that's Conversation AI (full build in Train Conversation AI). If the job is answering the phone, that's Voice AI. If you need one AI step inside a workflow you already have, the AI Agent action is simpler. The full map of every AI tool, with prices, is in GoHighLevel AI tools.

The way I explain any AI agent: "every AI agent has three components." A knowledge base, which is "the brain", what it knows. The prompt, the "instructions on how the AI agent should behave." And actions, "giving it the ability to do more than speak." Managed Agents use the same three pieces. The builder just writes most of the prompt for you.

Access, plans and the Labs toggles

Work in the client's sub-account, since the agent uses that account's contacts, forms and calendars. Your user needs permission to access and manage Managed Agents. Set up the knowledge base, forms, calendars and workflows the agent will use first, because the builder picks from what's already in the account.

Decide how the sub-account pays. Without an AI Employee plan, Managed Agents are pay per use from the wallet. Plans are switched on at the agency level under Agency View > Settings > Company > AI Employee, then per sub-account in Agency View > AI Suite > Plans & Limits.

Optional: in the sub-account's Labs, turn on "Managed Agents New UI" for the Unified Workspace that launched Oct 28, 2026. It puts the agent list, chat and test panel, and settings on one screen, and lets you test a trigger-based agent against a contact you pick. Without it you get the older screens, and the steps below still work.

What a Managed Agent costs to run

Managed Agents are billed by runs. HighLevel counts one run per chat session, one per trigger event (a tag added that fires the agent 12 times in a month is 12 runs) and one per scheduled run (daily at a fixed time is 30 runs in 30 days). Subagents are billed inside the parent's run. When an agent waits on a scheduled follow-up, the waiting is free and it's billed when it wakes up and works.

So size it before you sell it. The lead triage agent below runs once per form submission. A client with 60 web leads a month fits inside Growth's 100 runs. One with 400 needs Unlimited, or Extra Usage turned on in AI Suite, because when the allowance runs out the agent pauses by default. Rebilling AI usage to the client with a markup needs the $497 agency plan. Every other per-unit fee is in Usage fees.

Sub-account planManaged Agent runsFlow agents
No AI Employee planPay per use, by tokensPay per use
AI Employee Starter, $20 (only for agencies on the $497 plan)Not included, pay per usePay per use
AI Employee Growth, $50100 runs includedPay per use
AI Employee Unlimited, $971,000 runs includedPay per use
Testing (any plan)30 free test runs a month, up to 400,000 tokens per chat testPay per use

From HighLevel's AI Product Pricing doc (updated Nov 6, 2026) and gohighlevel.com/pricing, read Nov 9, 2026. Per sub-account per month.

Run a Managed Agent from a workflow

"Trigger agent studio from workflows" is one of the requests on HighLevel's ideas board, and for Managed Agents it's there now. Open the agent, add the Invoked from a workflow trigger and publish it. Then in Automation > Workflows add the Invoke Managed Agent action, pick the agent and write the instruction, with custom values for the contact and anything the workflow already knows. The agent's answer comes back as a value you can use in the next steps: an If/Else branch, an Update Contact Field, or the text of a message.

This is how I'd keep things predictable. The workflow owns the timing, waits and messages, and the agent only does the part that needs judgment. Flow agents have their own action, Invoke Agent Studio Agent, and the AI Agent action is the third option, built right inside the workflow. More on the builder itself in GoHighLevel workflows.

If software outside HighLevel needs the agent, the Invoked via API trigger gives you ready-made cURL, Python or JavaScript requests. API calls always hit the published version, and an unpublished agent returns a 404.

Triggers you can use

One agent can have several triggers. As of Nov 9, 2026 HighLevel's docs list:

  • Chat: you or your team talk to the agent directly.
  • Form Submitted and Survey Submitted: the agent gets the answers. A form's file upload can pass a PNG image to it.
  • Facebook lead form: pick the page and all or some lead forms. The page has to be connected for lead sync, and a page admin has to accept Meta's Lead Ads terms.
  • Tag Changed, Contact Created, Opportunity Created and Opportunity Status Changed (filter by pipeline, stage and Open, Won, Lost or Abandoned).
  • Appointment events and Schedule: once, every set interval, a cron expression, or just tell the builder "every weekday at 8am". Schedules run in the sub-account's time zone.
  • Invoked from a workflow and Invoked via API, covered above.

What's new and what's still beta (Oct 28, 2026 launches)

Computer: turn it on under the agent's Capabilities and the agent gets its own computer to browse websites, fill in forms, create files and run Python, with 8 to 10 browser chats in parallel. Logins are kept in HighLevel Keychain, and it asks before payments, sending messages, submitting forms or deleting files (those wait in Waiting on You). The catch: Computer works in chat and manual runs only, not runs started by a trigger, so the lead triage agent above can't browse on its own.

Subagents: the agent can split a big job into smaller tasks and run them in parallel, each with only the tools it needs. No setup, and the usage is billed inside the parent run. You can also tell the builder "Split this task into subagents."

n8n import, in beta: AI Agents > Agent Studio > Import from > n8n (Beta), upload the workflow JSON, and HighLevel shows which steps were Carried Over, Adapted or Not Possible. Credentials never come across, so reconnect every app before you publish.

Memory, which lets an agent remember user details and reuse steps it learned, is switched on through Labs. Scheduled follow-ups let an agent pause and pick back up 5 minutes to 30 days later, for things like a reminder before an appointment or a check three days after a sales call.

Copy the agent to other clients with a snapshot

Once an agent works in one account, don't rebuild it. Go to Agency View > Account Snapshots > Create New Snapshot, pick the source sub-account, expand Agent Studio, tick the agent, and also tick what it depends on: the knowledge base, the workflows it triggers and any custom values. Snapshots only carry what you select, so an agent without its knowledge base loads with no brain.

The snapshot keeps the instructions, triggers, skills and capabilities. Integrations, logins and other third-party connections need setting up again in each client account. When you improve the agent, edit it in the source account, click Refresh on the snapshot and push the update. Put the business details (name, service area, owner's name) in custom values so the same agent fits every client. How I structure niche snapshots is in GoHighLevel snapshots.

If you already have flow agents

Leave them running. HighLevel says existing flow agents and their triggers keep working, and you can still open them under AI Agents > Agent Studio > Flow Agents, edit nodes, test, save to Staging and promote to Production. HighLevel doesn't force a move, and its doc says to manage current flow agents in Agent Studio unless you have your own migration plan.

Two things to watch. Flow agents are pay per use on every plan, so an AI Employee plan doesn't cover them. And they can't be created any more, so build anything new as a Managed Agent and move the old ones over when you're editing them anyway.

How to sell this to a local business

Find the gap first. My advice for any speed-to-lead build: go on their site, text them after hours, fill out the contact form and see how long it takes for someone to get back to you. "Do they do it instantly? Do they take 6 hours, 24 hours?" If leads sit there, the triage agent plus a fast follow-up workflow is the fix.

Then sell the outcome: every lead is read, sorted and noted within a minute, and the owner only gets pinged for the jobs worth a call. Build it once, put it in your niche snapshot, and you could build this for a local business and charge a monthly fee you set. More on pitching it in Sell AI to local businesses.

Start the 30-day trial and build your first agent →
Start your 30-day free trial →

30 days free instead of HighLevel's usual 14, plus a 1:1 setup call with my team, weekly live calls and my agency resources from day one. See everything you get →

Other ways to start

My bonuses come with every one of these.

Best if you're new and want live help setting up

30-day trial + a free live setup bootcamp

Best if you're not sure which plan you need

Compare all 3 plans first

Best if you'll resell HighLevel as your own software

Go straight to the $497 Founder plan

FAQ

What is Agent Studio in GoHighLevel?

It's where HighLevel's AI agents are built and managed, under AI Agents > Agent Studio. New agents are built as Managed Agents by describing the job in plain English. The older flow-based canvas is still there for editing existing flow agents.

Is Agent Studio the same as AI Studio?

No. AI Studio builds websites, landing pages and small apps from a prompt. Agent Studio builds AI agents that act on CRM events. HighLevel's docs call them separate products.

What happened to Super Agents?

They were renamed Managed Agents. They live under AI Agents > Agent Studio > Managed Agents and are now the only way to create a new agent in Agent Studio.

How much do Managed Agents cost?

They're billed per run by tokens. AI Employee Growth ($50 a month per sub-account) includes 100 runs and Unlimited ($97) includes 1,000. Starter and no plan are pay per use. 30 test runs a month are free (as of Nov 9, 2026).

Can I still create flow-based agents?

No. HighLevel stopped new flow agents. Existing ones keep running, can still be edited and promoted from Staging to Production, and are billed pay per use on every plan.

Can a workflow trigger an Agent Studio agent?

Yes. Give the Managed Agent the Invoked from a workflow trigger, publish it, then add the Invoke Managed Agent action in the workflow. Its answer can drive later steps. Flow agents use the Invoke Agent Studio Agent action.

Can I share a Managed Agent with other sub-accounts?

Yes, through an account snapshot. Select the agent under Agent Studio plus the knowledge base, workflows and custom values it needs. Integrations and logins have to be reconnected in each account.

Sources / last checked Nov 9, 2026

  • HighLevel help docs: Managed Agents. Setup and Use Managed Agents in Agent Studio (modified Nov 8, 2026), Guided Managed Agent Builder (Oct 21, 2026), CRM triggers (Oct 21, 2026), Scheduled Triggers (Aug 12, 2026), Scheduled Follow-Ups (Oct 27, 2026), Activity Feed (Oct 8, 2026), Unpublish (Sep 10, 2026), Custom Skills (Sep 2, 2026), Memory (Oct 27, 2026) and Deploy to API (Oct 5, 2026), read Nov 9, 2026.
  • HighLevel help docs: Agent Studio and AI tools. Overview of Flow Agents in Agent Studio (modified Oct 6, 2026), AI Tools in HighLevel (Oct 27, 2026) and Managed Agents, Conversation AI & Ask AI: What's the Difference? (Oct 1, 2026), read Nov 9, 2026.
  • HighLevel help docs: workflows and snapshots. How to Invoke Managed Agents from HighLevel Workflows (modified Sep 24, 2026), Clone and Share Managed Agents Using Snapshots (Aug 12, 2026) and Clone and Share Agent Studio Agents Using Snapshots (Jun 26, 2026), read Nov 9, 2026.
  • HighLevel pricing. AI Product Pricing (modified Nov 6, 2026), AI Employee Access, Rebilling and Reselling (Oct 27, 2026) and gohighlevel.com/pricing (Growth $50, Unlimited $97 per sub-account per month; Starter $20 is listed in the help docs only), read Nov 9, 2026.
  • HighLevel changelog. Computer for Managed Agents, Subagents in Managed Agents, Import n8n Workflows into Managed Agents (Beta) and A Unified Workspace for Managed Agents, all Oct 28, 2026, read Nov 9, 2026. No newer Agent Studio entries through Nov 9, 2026.
  • My videos. "Build PERFECT AI Agents (NEW HACK)" (Aug 2026) for testing agents before they go live, and my dental speed-to-lead build (Jun 2026) for the three parts of an agent and the after-hours test. That one's title and opening have an income figure, so it isn't embedded.
Hamza Baig (Hamza Automates)

Author

Hamza Baig / Hamza Automates

Founder, Hexona Systems

I run Hexona Systems on GoHighLevel. HighLevel gave it the SaaSpreneur Platinum award in 2024 and Diamond in 2025 and 2026, and I spoke at their LevelUp event in 2026. Hexona powers 1,000+ agencies, we've served 1,500+ businesses, and I've taught 45,000+ students. Everything on this hub is something I use myself or have shown on camera.

Platinum SaaSpreneur
2024
Diamond SaaSpreneur
2025, 2026
LevelUp speaker
2026
YouTube subscribers, Oct 2026
34,200

Keep reading

Start your 30-day free trial →

30 days free + 1:1 setup call + my bonus / affiliate link