The short answer
GoHighLevel Agent Studio is where custom AI agents are built. Its original canvas builds flow agents from connected nodes (LLM, knowledge base, web search, API and MCP tools) with variables and routers, tested and promoted from Draft to Staging to Production. HighLevel now recommends Managed Agents, the conversational builder in the same area, for all new agents, and keeps the canvas for agents already built there. Agent Studio usage is pay per use on every plan, charged by tokens, API calls, searches or generated content. Managed Agent runs are included on AI Employee plans: 100 a month on Growth, 1,000 on Unlimited.
| Plans | All plans; flow agents are maintained here and new agents are built as Managed Agents |
|---|---|
| Cost | Pay per use on every plan; Managed Agent runs included: 100/mo on Growth, 1,000/mo on Unlimited |
| Checked | 2026-09-29, against HighLevel's help center and pricing page |
Flow agents on the canvas
A flow agent is drawn as a graph. HighLevel's overview lists the main node types, and related articles add a router tool for AI or conditional branching, capture tools for collecting user input, folders and snapshot cloning. HighLevel now positions the canvas for maintaining existing flow agents, so it suits agents already built there more than new projects. The building blocks:
- LLM node: generates responses with a large language model
- Knowledge Base tool node: answers from an attached HighLevel knowledge base
- Web Search tool node: pulls current information from the internet
- API tool node: sends and receives data from third-party services, with cURL import
- MCP tool node: connects external tools through the Model Context Protocol
- Global, runtime and input variables, plus Draft, Staging and Production versions and a Test tab
Managed Agents: describe the agent instead of drawing it
Managed Agents, formerly called Super Agents, are built by describing the agent's goal, the channels it should use, the knowledge bases it may read and the CRM actions it may take. HighLevel then generates the configuration. Triggers include tag changes, form and survey submissions, Facebook lead forms, a chat interface, and appointment, contact and opportunity events. Actions include updating contact fields and tags, triggering workflows, and creating tasks, notes or appointments, and agents can also generate images, audio and video. Agents move from Draft to Test to Published, and can be extended with MCP connectors and custom skills or copied between accounts with snapshots.
Running agents from workflows
The Invoke Agent Studio Agent workflow action runs a flow agent in the middle of a workflow and stores its response for later steps, such as an email, an SMS, a branch or a field update. The agent must be published to Production and built with the Workflows trigger, and it must answer within 60 seconds. HighLevel's article, updated May 2026, lists an active AI Employee add-on as a requirement for this action and says runs use the same metering as in Agent Studio, with no workflow surcharge. A separate Invoke Managed Agent action does the same for Managed Agents. Typical uses are lead enrichment, call summaries, sales triage and structured JSON for branching.
What it costs
HighLevel's AI pricing article puts Agent Studio on pay per use for every plan, with charges based on token usage, API calls, searches or content generation units depending on the node. There is no flat allowance for flow agents in AI Employee. Managed Agents are counted in runs, one run per agent execution: AI Employee Growth ($50 a month per sub-account) includes 100 runs and Unlimited ($97) includes 1,000, with pay per use beyond that. Since September 23, 2026, GPT-6 Sol and GPT-6 Luna are available in the AI Agent and GPT workflow actions. Watch agent logs during the first week, because a loop in a flow can run up token costs quietly.
When to use it instead of Conversation AI
Conversation AI and Voice AI cover the common jobs: answering leads, qualifying them and booking appointments, configured through forms rather than graphs. Agent Studio earns its place when an agent needs to call an outside API, search the web, route between specialist sub-agents or return structured data to a workflow, for example scoring a lead against a client's own pricing system. For agencies, the practical rule is to build a Managed Agent first, since it is HighLevel's recommended path, and only drop to a flow agent when a step needs exact control that a description cannot give.
Costs add up differently for every account. Put your numbers into the HighLevel cost calculator, or browse every HighLevel feature.
Cost your account Try HighLevel