GoHighLevel Workflows API and Workflows in the LeadConnector MCP Server: Ask Claude What Broke in Your Workflow

Status check: both features are private beta, per HighLevel's changelog. The Workflows API is enabled for selected sub-accounts only, and the MCP workflow tools only work in sub-accounts enabled for the Workflows API. Every other sub-account gets the error WORKFLOW_API_NOT_ENABLED_FOR_LOCATION. HighLevel's LevelUp Day 2026 roundup lists the Workflows API (private beta); the MCP entry is on the changelog. It isn't in my account yet, so this is based on HighLevel's release notes, not my own testing.

Anyone who manages a lot of GoHighLevel workflows knows the slow part is not building them. It's finding out why one stopped working, for which contacts, and at which step. That's the angle I care about here: asking an AI assistant what broke instead of clicking through execution logs one workflow at a time.

The two features side by side

Ask Claude what broke in your workflow

From the MCP changelog entry, an assistant connected to the LeadConnector MCP server can:

• Find workflows by name, status, folder or date.
• Read a workflow's full definition: triggers, steps, branches and settings.
• Build a workflow from a description. The assistant checks it for errors and saves it as a draft.
• Update and publish a workflow, or set it back to draft.
• Delete a workflow, with an extra confirmation first.
• Read execution logs: who entered a workflow, what happened at each step and which steps failed, with the error.
• Count statistics: enrollments, completions and failures per step, by hour, day or week.

How to connect your AI tool (steps from HighLevel)

1. Make sure your sub-account is in the Workflows API private beta (contact HighLevel support to join).
2. Add the LeadConnector MCP server to your AI tool:
   • Claude: Settings > Connectors > Add custom connector, then enter https://services.leadconnectorhq.com/mcp/anthropic/v2
   • ChatGPT: turn on Developer mode, open ChatGPT Plugins and add https://services.leadconnectorhq.com/mcp/openai/v2/ as the connection
   • Cursor: use https://services.leadconnectorhq.com/mcp/cursor/v2
3. Sign in with your LeadConnector account, choose the sub-account and approve the workflow permissions.
4. Start a new chat and ask about your workflows. HighLevel says the LeadConnector MCP server docs have setup steps for every client.

Prompts I would start with

HighLevel's examples are failure triage, weekly numbers and drafting by description. Here's how I'd phrase them for a client account:

• "Which steps in the New Lead Nurture workflow failed this week? List each failed step with the contact and the error."
• "How many contacts entered the Webinar Follow-up workflow each week this month, and how many completed it?"
• "Which step in the Missed Call Text Back workflow has the most failures in the last 7 days?"
• "Draft a workflow that tags new contacts webinar-lead and sends a welcome SMS. Save it as a draft, don't publish it."

The first two are read-only, which is where I would start. Reading logs and counts can't break anything. Building and publishing can.

The Workflows API for your own code

If you'd rather build your own tools, the Workflows API changelog entry lists:

• Create workflows from a complete definition. New workflows start as drafts.
• Read and update workflows as a definition you can change and send back.
• Validate before saving: check a definition for errors without saving, with each error pointing at the exact step.
• Publish or unpublish a workflow.
• Search and delete by name, status, folder or date.
• Step types: browse every trigger and action available in the sub-account, with its fields.
• Execution logs and statistics, the same as in MCP.

HighLevel's LevelUp Day roundup adds that the API has ten new endpoints and uses a consistent definition format.

How to get started with the API (steps from HighLevel)

1. Contact HighLevel support to have the private beta enabled for your chosen sub-accounts.
2. In the sub-account, go to Settings > Private Integrations, create an integration and select the workflow scopes you need: View Workflows, Edit Workflows, Publish Workflows, Delete Workflows and View Workflow Logs.
3. Send the token as a bearer token, with the Version: v3 header and the sub-account's locationId on every request.
4. Use the Workflows API reference for every endpoint and its fields. HighLevel says each endpoint page has a request panel to try it with your token.

New to private integration tokens? My Private Integration Token setup guide walks through creating one. Only tick the scopes you need: for a reporting tool, View Workflows and View Workflow Logs are enough.

Use cases HighLevel gives

• Standard rollout: read a workflow from your template sub-account, create it in each client sub-account, then publish it.
• Failure report: every Monday, count failures per step for your key workflows, list who was affected and send the summary to your team.
• Onboarding from your app: a client signs up in your product, and your app creates and publishes their onboarding workflow in their sub-account.

Read this before you let an AI edit live workflows

These notes come straight from HighLevel's changelog and they matter more than the features:

• Editing a published workflow pauses it. It has to be set to draft first, and while it's a draft its triggers are off and contacts in it are paused.
• Updates replace the whole workflow (API), so read it first and send back the version you read.
• Deleting is permanent and stops everyone currently in the workflow.
• New workflows are drafts. Publishing is a separate request.
• One sub-account per request in MCP. HighLevel says agency connections that cover several sub-accounts are rolling out.
• App steps only accept an app account that's connected in the sub-account.
• Logs cover up to 30 days per request and go back 60 days. They show live runs only; test runs from the builder aren't included.

Here's how I would set it up for a client: connect with read-only questions first, build any new workflow as a draft and review it in the builder, and only make changes to a live workflow outside busy hours, because contacts in it pause while it's a draft.

Where this fits with Workflow Maps

HighLevel also announced Workflow Maps this week, with a workflow errors view and contact journeys inside the app. I covered that in my Workflow Maps guide. The MCP and API route is for when you want answers in your own AI tool or your own reports instead.

For the rest of this week's HighLevel news, see my running LevelUp 2026 roundup.