Keel is WrightLabs' control layer between AI and GoHighLevel. It's the protocol that lets Claude, ChatGPT, Gemini, or any MCP-compatible model read, write, route, and act inside any GHL location — your own, your clients', or your partners'. Named after the part of the boat nobody sees but everything depends on.

How is Keel different from GHL's built-in AI?

GHL's built-in AI runs inside workflow steps. Keel lets you command the entire platform interactively — across contacts, pipelines, conversations, social, payments, and more — in a single conversation. It also runs scheduled autonomous agents and supports cross-sub-account orchestration.

What AI models does it work with?

Currently Claude Desktop and Claude Code. Expanding to GPT and Gemini compatibility. Any MCP-compatible client will work because Keel speaks the open Model Context Protocol standard.

Is my GHL data secure?

Install is OAuth-only via the GHL Marketplace. WrightLabs does not store credentials. All actions are sub-account scoped and auditable through Keel's replay log.

What GHL plan do I need?

Any GHL plan with API access. Works with sub-accounts on Agency Unlimited and Pro/SaaS Mode plans.

Can it operate multiple sub-accounts?

Yes. Multi-sub-account orchestration is a core capability. Deploy workflows across all your sub-accounts, run cross-account reports, or replicate winning configurations from one location to thirty others in a single command.

// Keel by WrightLabs · Free install via GHL Marketplace

The part of the boat nobody sees — but everything depends on.

Keel is the control layer between AI and your GoHighLevel. It's the protocol that lets Claude, ChatGPT, Gemini, or any model read, write, route, and act inside any GHL location — your own, your clients', your downline's. One layer. One discipline. Every sub-account.

Install Keel Start a 20-min build call

claude — keel

// One layer over your entire GHL footprint

$ claude "Move all stale Q2 opps to nurture and fire the reactivation workflow on each one"

→ Reading 847 opportunities across your connected accounts...

→ 312 qualify (no activity 14+ days)

→ Moving pipeline stages... ✓

→ Triggering reactivation workflow... ✓ 312 contacts

→ ETA first responses: ~4 hours

Done in: 12 seconds.

Manual equiv: ~6 hours.

// What it actually does

Not a query tool. An operator.

Most GHL integrations let you read contacts. Keel lets you run the platform — every surface, every action, every sub-account you operate — through natural language. With the guardrails of deterministic compliance and the audit trail of a replay log.

Contacts & CRM

31 tools

Create, search, merge, tag, DND, bulk update

Conversations

20 tools

SMS, email, WhatsApp — send, read, assign, schedule

Pipelines & Deals

10 tools

Stage moves, deal values, pipeline velocity

Calendar & Bookings

14 tools

Book, reschedule, check slots, assign staff

Social Media

17 tools

Post to FB/IG/LinkedIn/X/GMB natively from chat

Payments & Billing

20 tools

Transactions, subscriptions, payment links, invoices

Store & Commerce

18 tools

Products, orders, inventory, coupons, fulfillment

Campaigns & Workflows

Full CRUD

Create, trigger, clone, publish, disable, monitor

Snapshots & Config

24 tools

Export, deploy, diff, replicate across sub-accounts

Reporting & Analytics

Full coverage

Pull KPIs, dashboards, performance data on demand

+ 30 more categories including forms, surveys, funnels, blogs, phone system, voice AI, LMS, email infrastructure, triggers, social OAuth

// The part no other GHL integration has

Add the WrightLabs AI front desk. Close the loop from lead to payment.

Pair the MCP with the WrightLabs AI front desk and the entire outbound revenue motion runs itself. Lead enters GHL → AI voice agent calls within 60 seconds → books the appointment → sends the payment link → updates the CRM → reconfigures the agent prompt based on conversion data. One prompt. No humans required.

01 Lead enters

02 Voice agent dials

03 Books appointment

04 Sends payment link

05 CRM updated

06 Agent self-optimizes

The agent also reconfigures itself. When conversion drops, the MCP rewrites the agent's global prompt, swaps tools, and redeploys live. Performance-triggered prompt evolution — running while you sleep.

// Installation

Connected in 5 minutes.

STEP 01 Visit wrightlabs.ai/mcp → click Install No account creation. No payment up front. OAuth handshake initiates from this page.

STEP 02 Authorize via GHL Marketplace OAuth Sub-account level. All required scopes. Standard GHL Marketplace flow — fully auditable.

STEP 03 Add server to Claude Desktop or Claude Code One config block. We provide the JSON. Start commanding your GHL in natural language.

Works with Claude Desktop, Claude Code, and any MCP-compatible AI client.

// Scheduled · autonomous

Set it once. It runs forever.

The MCP isn't just interactive — it's schedulable. Daily pipeline cleanup at 6am. Hourly lead routing. Weekly performance reports auto-generated and emailed. Monthly billing reconciliation. Cross-account audits that flag drift before your clients notice. You review the output. You don't generate it.

scheduler.wrightlabs.ai

06:00 daily → pipeline_cleanup --all-accounts

every 1hr → lead_router --round-robin --fallback=5m

Mon 08:00 → performance_report --email-agency-owner

01st monthly → billing_reconcile --flag-failed-payments

Sun 22:00 → snapshot_diff --compare-to-gold-standard

The part of the boat nobody sees — but everything depends on.

Not a query tool. An operator.

Add the WrightLabs AI front desk. Close the loop from lead to payment.

Connected in 5 minutes.

Set it once. It runs forever.

The short answers.