1. Home
  2. Nexus AI
  3. Nexus +
  4. AI Assistants MCP – User Guide
  1. Home
  2. Nexus AI
  3. AI Assistants MCP – User Guide

AI Assistants MCP – User Guide

Launch365.ai can connect to ChatGPT, Claude, and other MCP-compatible assistants through the Model Context Protocol, or MCP. This lets an approved assistant answer questions about your Launch365.ai company data and create reviewable drafts or approved updates without sharing your Launch365.ai password or third-party integration credentials.

Use this guide to turn on the AI Assistants integration, connect an MCP client, choose permissions, and start asking useful questions.


What MCP Does in Launch365.ai

MCP gives an outside AI assistant a controlled way to use Launch365.ai tools on your behalf.

With the integration connected, an assistant can help with work such as:

  • finding contacts and summarizing customer history
  • reviewing recent conversations and contact activity
  • drafting SMS replies without sending them
  • creating follow-up tasks
  • reviewing opportunities, estimates, invoices, jobs, and appointments
  • summarizing marketing performance and Analytics & Insights reports
  • summarizing connected Google reporting data
  • listing campaigns and creating campaign drafts
  • reviewing customer reviews and drafting owner responses
  • reading and updating custom fields when allowed
  • applying tags when allowed
  • suggesting FieldOps booking, scheduling, and technician actions when enabled

MCP is company-scoped. The assistant can only access data for the Launch365.ai company connected by the signed-in user.


Before You Start

You need:

  • Nexus Plus enabled for the Launch365.ai company.
  • A Launch365.ai user login that can access the company and the Integrations area.
  • An AI assistant or MCP client that supports remote MCP servers over HTTP with OAuth.
  • Any source integrations already connected if you want to ask about that data, such as Google Analytics, Google Search Console, Google Ads, Google Business Profile, or FieldOps.

Do not connect Launch365.ai MCP to a shared assistant account. Use the individual assistant account for the person who should have access.


Turn On the AI Assistants Integration

  1. Sign in to Launch365.ai.
  2. Go to Settings -> Integrations -> AI Assistants.
  3. The direct app path is /company/integrations/ai-assistants.
  4. Turn on Enable AI assistant connections.
  5. Choose the capability groups the assistant should be allowed to use.
  6. Leave Require confirmation before write actions enabled unless you have a specific reason to turn it off.
  7. Click Save Settings.
  8. Copy the MCP server URL shown in the Connection panel.

The production MCP server URL usually looks like:

https://go.launchsms.com/mcp

Always copy the URL from the Launch365.ai integration page when setting up a non-production environment.


Connect an MCP Assistant

The exact screen names vary by assistant, but the connection flow is usually the same.

  1. Open the assistant’s connector, integrations, or MCP settings.
  2. Add a custom MCP connector or remote MCP server.
  3. Paste the Launch365.ai MCP server URL.
  4. Save or continue.
  5. When the assistant opens the Launch365.ai login and authorization screen, sign in.
  6. Review the requested permissions.
  7. Approve the connection.
  8. Return to the assistant and test the connection with a simple question.

Good first test prompts:

  • Find my most recently updated contacts.
  • Summarize recent customer conversations.
  • What Launch365.ai tools do you have access to?
  • Show me marketing performance for the last 30 days.

If the assistant asks for technical metadata URLs, use these production URLs:

Protected resource metadata:
https://go.launchsms.com/.well-known/oauth-protected-resource/mcp

Authorization server metadata:
https://go.launchsms.com/.well-known/oauth-authorization-server

For non-production environments, use the same paths on the domain shown inside Launch365.ai.


Common Ways to Use It

Contacts and Conversations

Ask the assistant to find people, summarize relationship history, and prepare replies.

Example prompts:

  • Find John Smith and summarize his latest activity.
  • Show me recent conversations that may need a reply.
  • Draft a friendly SMS follow-up for customer 123 about their estimate.
  • What notes and SMS activity do we have for this contact?

Drafted SMS replies are saved as drafts only. Launch365.ai MCP does not send the message directly.

Tasks and Follow-Up

Use MCP to find missing follow-up and create internal tasks.

Example prompts:

  • Show me open follow-up tasks due this week.
  • Find hot leads without a follow-up task.
  • Create a follow-up task for customer 123 for tomorrow at 9 AM.

If write confirmation is enabled, the assistant must confirm the action before creating the task.

Marketing and Reporting

Use MCP to review campaigns, automations, and Analytics & Insights reports.

Example prompts:

  • Which campaigns need attention this month?
  • Which automations are underperforming in the last 30 days?
  • Show Email Performance for the last 30 days.
  • Show Quote Wizard Analytics for the last 90 days.
  • Create a draft SMS campaign for open estimates from last month.

Campaign creation through MCP saves a draft only. It does not send or schedule the campaign.

Google Reporting

If Google integrations are connected, MCP can summarize performance without exposing Google OAuth tokens.

Example prompts:

  • List the connected Google integrations.
  • Summarize Google Business Profile performance this month.
  • How did Google Search Console perform over the last 30 days?
  • Show Google Ads spend, clicks, and conversions this month.
  • Summarize Google Local Services Ads lead performance.

If a Google report is empty, confirm that the matching Google integration is connected and has recent data.

Reviews

Use MCP to review feedback and draft owner responses.

Example prompts:

  • List recent 5-star reviews.
  • Show reviews that do not have replies.
  • Draft a professional reply to this review.

Review replies are returned as drafts only. MCP does not post owner responses.

Opportunities, Estimates, Invoices, Jobs, and Appointments

Use MCP to inspect pipeline and service history.

Example prompts:

  • Show open opportunities from the last 30 days.
  • List estimates that may need follow-up.
  • Show outstanding invoices.
  • List jobs scheduled for next week.
  • Show appointments for customer 123.

When allowed, MCP can update opportunity status fields. It does not send customer messages as part of that update.

Custom Fields, Tags, and Notes

Use these actions when your team wants the assistant to organize customer records.

Example prompts:

  • List custom fields for this account.
  • Update the Service Type custom field for customer 123 to HVAC.
  • Apply the Hot Lead tag to customer 123.
  • Create an internal note on customer 123 summarizing this conversation.

These are write actions. Keep confirmation enabled so the assistant has to review the change before it is saved.

FieldOps

FieldOps MCP tools are optional and are not part of the default permission set. Enable the FieldOps read/write capability groups only for teams that want assistant help with dispatch workflows.

Example prompts:

  • Show pending FieldOps bookings.
  • Suggest whether this booking should be converted to a job.
  • Find schedule availability for next Tuesday.
  • Suggest the best technician for this job.
  • Schedule this job for the recommended slot.

FieldOps write actions may also require Nexus Autopilot confirmation in FieldOps settings.


Permission Groups

Capability groups control what tools the assistant can see and use. If a tool is missing in the assistant, the matching permission group may be disabled or the assistant may need to reconnect after permissions changed.

Permission group What it allows
contacts.read Read contacts and customer profiles.
conversations.read Read conversation history and contact activity.
reports.read Read marketing reports and automation performance.
analytics.read Read Analytics & Insights reports.
messages.draft Create SMS reply drafts without sending.
tasks.read Read follow-up tasks.
tasks.write Create follow-up tasks.
opportunities.read Read opportunity and pipeline records.
opportunities.write Update opportunity status and related pipeline fields.
estimates.read Read customer estimates.
invoices.read Read customer invoices.
jobs.read Read jobs and service history.
appointments.read Read appointments and scheduling context.
campaigns.read Read SMS broadcasts and email campaigns.
campaigns.draft Create draft campaigns without sending or scheduling.
reviews.read Read customer reviews.
reviews.reply_draft Draft owner responses without posting.
custom_fields.read Read custom fields and customer field values.
custom_fields.write Update customer custom field values.
company.read Read safe company profile context, business hours, public phone numbers, service area, timezone, and connected integration names.
google.integrations.read Read connected Google integration status.
google.business_profile.read Read Google Business Profile summary data.
google.analytics.read Read Google Analytics summary data.
google.search_console.read Read Google Search Console summary data.
google.ads.read Read Google Ads summary data.
google.local_services_ads.read Read Google Local Services Ads summary data.
notes.write Create notes on customer records.
tags.write Apply tags and lead status updates.
fieldops.read Read FieldOps pending bookings, schedule, and technician workload.
fieldops.write Create bookings, convert bookings to jobs, schedule jobs, and assign technicians.

Recommended setup:

  • Start with read-only groups plus draft groups.
  • Add write groups only for users who should make changes from an assistant.
  • Keep fieldops.writecustom_fields.writeopportunities.writenotes.write, and tags.write limited to trusted users.
  • Reconnect the assistant after changing permission groups so the OAuth connection can request the updated scopes.

Write Action Safety

Launch365.ai separates read actions, drafts, and write actions.

Read actions can return information from Launch365.ai, such as contacts, conversations, reports, and connected integration summaries.

Draft actions create or return draft content only. Examples include SMS reply drafts, campaign drafts, and review reply drafts. Draft actions do not send messages or post replies.

Write actions change Launch365.ai records. Examples include creating a task, updating an opportunity, updating a custom field, creating a note, applying a tag, or scheduling FieldOps work.

When Require confirmation before write actions is enabled, the assistant must include explicit confirmation before the write action runs. If confirmation is missing, Launch365.ai rejects the write action.


Security Notes

  • Launch365.ai uses OAuth for MCP connections. Do not paste your Launch365.ai password into an assistant chat.
  • Google and other third-party OAuth tokens stay inside Launch365.ai. MCP responses return scoped report data, not raw credentials.
  • MCP tools are scoped to the authenticated user’s company.
  • The assistant can only use tools for capability groups enabled in Launch365.ai and approved during OAuth.
  • Use the smallest permission set that supports the user’s work.
  • Keep write confirmation enabled for normal use.
  • Remove the connector inside the assistant when a user should no longer have access.
  • To pause company-wide MCP access, turn off Enable AI assistant connections and save the integration settings.

Troubleshooting

The assistant says it is not authorized

Reconnect the Launch365.ai connector from the assistant and complete the Launch365.ai OAuth approval flow.

The assistant says Nexus Plus is required

The company must have Nexus Plus enabled before MCP tools can be used.

The assistant says the integration is disabled

Go to Settings -> Integrations -> AI Assistants, turn on Enable AI assistant connections, and save.

A tool or data type is missing

Check the allowed capability groups on the AI Assistants integration page. Enable the needed group, save, then reconnect the assistant so it can request the updated scope.

A write action is rejected

If confirmation is enabled, the assistant must confirm the change before Launch365.ai will save it. Ask the assistant to show the proposed change, then tell it to proceed.

Google reporting is empty

Confirm the matching Google integration is connected in Launch365.ai and has data for the requested date range.

FieldOps write actions are blocked

Confirm that fieldops.write is enabled for AI Assistants and that FieldOps Nexus Autopilot settings allow the requested action.

The assistant is showing the wrong company data

Disconnect the connector from the assistant, reconnect it, and sign in with the correct Launch365.ai user.


Best Practices

  • Ask specific questions with date ranges, customer names, or customer IDs.
  • Tell the assistant whether you want a summary, a list, a draft, or an action.
  • Review drafts before sending or posting anything.
  • For write actions, ask the assistant to show exactly what will be changed before confirming.
  • Keep high-risk write permissions disabled until your team has a clear workflow.

Example strong prompts:

  • Find contacts tagged Hot Lead with no open follow-up task, then suggest the next action for each.
  • Summarize webchat analytics and form analytics for the last 30 days.
  • Draft an SMS follow-up for open estimates from the last 14 days, but do not send anything.
  • List Google Ads campaigns with high spend and low conversions this month.
  • Show pending FieldOps bookings and recommend which should be converted to jobs.
Updated on June 28, 2026
Was this article helpful?

Related Articles

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.