ai-integration

Ask AI + MCP Server Integration Tutorial | Step-by-Step Setup Guide for the Platform

Charles Higgins

Charles Higgins

— 11 min read
Featured
Featured

Table of Contents

Introduction — Why this integration matters

We recently shared a walkthrough with Andrew George that demonstrates a major update for the platform: Ask AI now integrates directly with the MCP server (Model Context Protocol server). In plain terms, this means the integrated AI can safely and securely act inside your account—creating contacts, moving deals, retrieving account-level metrics, and much more—without leaving the platform or forcing your team to switch between tools.

This article expands on that video and turns the demo into a practical, step-by-step guide. We’ll explain what the MCP server does, how Ask AI uses it, real-world examples you can start using today, and best practices to onboard team members and clients. Our goal is to make the change simple, useful, and immediately valuable for busy teams and small business owners.

Outline

  • What the MCP server is and why it matters
  • How Ask AI connects to the MCP server
  • Live examples: creating contacts, moving pipeline opportunities, and retrieving revenue
  • Step-by-step usage and security safeguards
  • Brand voices, feedback, credits, and billing basics
  • Best practices, suggested workflows, and onboarding tips
  • Troubleshooting and common questions
  • FAQ

What is the MCP server (Model Context Protocol) and why it matters

The MCP server is the piece that allows AI language models to act like a user inside your platform account. Rather than being limited to generating suggestions or draft text, the integrated AI can perform actions directly—create contacts, update pipeline stages, pull account-wide reports, and more—while respecting the platform’s permissions and security model.

Think of the MCP server as a secure bridge: it gives Ask AI context and safe access to the data and actions inside your account without exposing credentials or requiring you to switch to a separate AI tool. For teams, the benefits are clear:

  • Save time: perform routine tasks with natural language instead of navigating multiple screens
  • Reduce context switching: keep all work inside your business software instead of jumping to external chat tools
  • Consistent voice and output: use brand voices so responses and content match your company tone
  • Automation-friendly: the AI can execute steps that previously required manual clicks, enabling faster workflows

How Ask AI integrates with the MCP server

Ask AI is the platform’s integrated assistant. When we invoke Ask AI from the top toolbar, it can either provide conversational answers or, when asked, request the MCP server to perform actions inside a subaccount. The MCP server grants restricted, auditable access so actions happen as if a user did them, but the source is AI-driven.

Key points about the integration:

  • The Ask AI button is accessible from the top-right of the platform across screens—so you can call on the AI no matter where you are.
  • Actions request the MCP server to carry out changes in the subaccount. The server ensures the right permissions and logs actions.
  • We can retrieve account-level information (not just contact-specific data) via natural language queries.
  • For irreversible or sensitive changes, the system requests an explicit confirmation before proceeding.

Real examples: what Ask AI + MCP server can do today

Watch the integration in action and you’ll see these core capabilities. Below we replicate the examples we used in the demo and expand on how you can use them in your day-to-day operations.

1. Create a new contact instantly

Scenario: A team member has a lead on the phone and wants to add them without navigating to the contacts screen. Instead of switching screens, we can open Ask AI and say:

"Create a contact named Jamie Smith with phone number 555-123-4567."

What happens:

  • Ask AI interprets the request and sends it to the MCP server.
  • The MCP server creates the contact in the subaccount with the required fields.
  • Ask AI confirms the contact has been created and we can refresh to see it live in the contacts list.

Why this is useful: this saves time during calls, reduces the risk of missed entries, and keeps notes in the right place immediately.

2. Move opportunities in a pipeline

Scenario: We want to move a lead from "prospect" to "customer" without opening the pipeline board.

We simply tell Ask AI:

"Move our opportunity in this pipeline to 'customer'."

What happens:

  • Ask AI identifies the opportunity and the pipeline context, then instructs the MCP server to update the opportunity stage.
  • Within moments, the pipeline updates and the opportunity appears in the new stage.

Why this is useful: field reps and account managers can update pipeline status on the go without learning extra navigation steps. It also reduces human error and accelerates reporting.

3. Retrieve account-level metrics conversationally

Scenario: We want a quick sense of revenue in open opportunities across a location or account without building a report.

Ask AI can answer queries like:

"How much revenue do we have in open opportunities?"

What happens:

  • The MCP server aggregates the account data and returns a conversational response—e.g., "A little over $1,000,000 in open deals."
  • Ask AI can follow up offered options, such as whether we want a breakdown by pipeline or stage.

Why this is useful: executives and managers can get quick situational awareness and take immediate actions or request follow-ups directly in the conversation.

Step-by-step: How to use Ask AI with the MCP server

The interaction is straightforward. Below is a practical flow you can adopt immediately.

Step 1 — Open Ask AI

  • Click the Ask AI button in the top right of the platform. This is available on most screens so you can act from anywhere.

Step 2 — Phrase your request

Keep prompts simple and explicit, especially for actions. For example:

  • Create contact: "Create a contact named [Full Name] with phone [number] and email [email]."
  • Move pipeline: "Move opportunity [Opportunity Name or Contact Name] to [Stage Name] in [Pipeline]."
  • Retrieve metrics: "How much revenue is currently in open opportunities for [Location or Account]?"

We find that inclusion of key identifiers (names, phone numbers, pipeline names) reduces ambiguity and improves speed.

Step 3 — Review and confirm

If the action is reversible and routine, Ask AI will execute it through the MCP server and give a confirmation. For irreversible or sensitive tasks, Ask AI will prompt for confirmation:

  • "Are you sure you want to delete [X]?"
  • We confirm or cancel as needed.

Step 4 — Validate the result

After execution, refresh the relevant screen or check the confirmation message to ensure the change went through. The MCP server logs the change so you can audit it if needed.

Safety, permissions, and auditability

Security is built into the flow. The MCP server acts as a controlled gateway that honors the platform’s permission model. Some important points:

  • Actions are scoped to a subaccount: the AI only has access to the data for the account in which it’s invoked.
  • Irreversible actions require explicit confirmation to prevent accidental changes.
  • We can thumbs up or thumbs down responses to give feedback and help the assistant improve.
  • Actions that complete successfully consume one AI credit; failed attempts do not charge credits.
  • All actions are logged so teams can check who or what made changes and when.

Brand voices, feedback, and AI credits

One of the major advantages is consistent communication. Ask AI uses brand voices that you define in the platform’s marketing settings. This allows the assistant to craft responses and content that sound like your business.

How to set up a brand voice:

  • Go to the marketing section and select brand boards (or the equivalent area in your software).
  • Create a brand voice by defining tone, vocabulary, and example phrasing.
  • Set a default brand voice so Ask AI uses it automatically for responses and content generation.

Feedback helps improve accuracy. At the end of an Ask AI response, use the thumbs up or thumbs down to indicate whether the output met your expectations. This is a simple way to help refine the assistant’s behavior over time.

Credits and billing basics:

  • Ask AI consumes one AI credit per successful action completion. If an action fails or Ask AI cannot complete the task, it does not deduct a credit.
  • This billing model gives us budget predictability while enabling frequent use for routine tasks.
  • We recommend monitoring credit usage initially to set reasonable usage policies for staff accounts.

Practical use cases and workflows

Here are concrete workflows that demonstrate how this integration reduces friction and saves time.

Onboarding new leads during calls

  • When a lead calls, the rep opens Ask AI and creates the contact using a single natural language phrase.
  • Ask AI then adds the contact to the appropriate pipeline and stage, sets a task reminder, and notes the source—without the rep leaving the call.

Quick pipeline updates for field teams

  • Field teams can update deal stages or record onsite outcomes via Ask AI on mobile (mobile support coming later), speeding up reporting and improving data freshness.

Executive check-ins without building reports

  • Managers can ask for account summaries such as "total open revenue" or "opportunities closing this month" and get instant, conversational answers.
  • Follow-up questions are conversational—e.g., "Can I get a breakdown by stage?"—and Ask AI will return more detailed data.

Automating routine account maintenance

  • Use Ask AI to batch-create contacts, update bulk fields, or schedule follow-ups based on natural language instructions, freeing admins from repetitive tasks.

To get the most value, follow these practical recommendations we use internally.

1. Define clear brand voices

  • Create brand voices that reflect how you want communications to sound: friendly, professional, or highly authoritative—whatever aligns with your brand.
  • Provide example phrases so the AI learns the tone you expect for emails, SMS, and chat responses.

2. Start with guided tasks for customers

When rolling this out to clients or team members, provide a short list of recommended commands that produce high-value results. For example:

  • "Create a contact with name, phone, and email"
  • "Move opportunity [X] to [Stage]"
  • "How much revenue do we have in open opportunities?"

These guided prompts reduce confusion and highlight immediate benefits.

3. Test everything in a safe environment

  • Before enabling AI actions widely, test commands in a sandbox or a test subaccount to ensure behavior matches expectations.
  • Verify edge cases such as duplicate contacts or ambiguous pipeline names.

4. Set usage policies and monitor credits

  • Decide who can perform actions through Ask AI and which actions are allowed to prevent misuse or accidental changes.
  • Monitor credit usage and adjust as necessary to manage costs.

5. Keep humans in the loop for sensitive actions

  • For financial changes, mass deletions, or anything with compliance implications, require a human confirmation or limit access to senior admins.

Mobile and future improvements

Mobile support for the Ask AI + MCP server integration is coming later in the year. When mobile arrives, field teams and on-the-go staff will be able to use the same natural language capabilities directly from their phones, making it even easier to add contacts, update opportunities, and check metrics from the job site or client visits.

We expect ongoing improvements around:

  • Broader action coverage (more tasks the AI can safely perform)
  • Better context handling for multi-step workflows
  • Enhanced integrations with brand voice settings and messaging templates

Troubleshooting and tips

If you encounter unexpected results, here are the most common checks and fixes:

1. Ambiguous commands

  • Problem: Ask AI asks for clarification or completes the wrong action.
  • Fix: Use more specific identifiers—names, pipeline names, or contact phone numbers—to reduce ambiguity.

2. Permission issues

  • Problem: The AI cannot perform an action because of permissions.
  • Fix: Confirm the user account has the necessary rights in the subaccount or assign AI usage to an admin-level role where appropriate.

3. Unexpected billing (credits)

  • Problem: Credits were spent but the action didn’t complete as expected.
  • Fix: Credits are only charged for successful completions. If something went wrong, check the confirmation logs and consider resetting the test and trying again. Use thumbs down feedback to flag failures so the system can improve.

4. Data not appearing immediately

  • Problem: After Ask AI confirms execution, the UI doesn’t show changes.
  • Fix: Refresh the page or the specific table view. The MCP server executes changes server-side; sometimes a UI refresh is needed to view the updated data.

How to onboard team members and clients

Successful adoption is about clarity and trust. Here’s a short playbook we use:

  • Start with a short training session demonstrating 3-5 core commands (create contact, move pipeline, get revenue).
  • Provide a one-page cheat sheet with recommended prompts and examples tailored for each role (sales, support, ops).
  • Set expectations about security and credits so teams know when to escalate actions to admins.
  • Collect feedback using the thumbs up/down mechanism and iterate on brand voices and recommended prompts.

Key limitations to understand

While this integration significantly extends the assistant’s capabilities, there are a few limitations to be aware of:

  • Not all platform actions are available through the MCP server yet—review current supported actions to design workflows appropriately.
  • Complex multi-step workflows may require more explicit instructions or follow-ups to ensure accuracy.
  • Data access is scoped to the subaccount in which Ask AI is used—AI won't cross account boundaries unless explicitly set up that way.

Example prompts to use right away

Below are prompts we recommend to produce reliable, high-value results:

  • Create a contact: "Create contact [Full Name], phone [number], email [email], source [lead source]."
  • Add note to contact: "Add note to [Contact Name]: [Summary of conversation]."
  • Update opportunity: "Move opportunity for [Contact Name] to [Stage Name] in [Pipeline]."
  • Get metrics: "How much revenue do we have in open opportunities for [Location]?"
  • Bulk action (example): "Create contacts for these names and phone numbers: [List]."

Real feedback and next steps

We’ve already heard excitement from teams that this reduces repetitive tasks and keeps work centralized. As one member put it during testing: "This is so cool — it moved the opportunity all on its own." That kind of reaction captures the core value: less manual work, fewer mistakes, and faster insight.

Our recommended next steps for teams interested in adopting Ask AI with the MCP server:

  1. Identify 3-5 highest-impact actions your team performs daily (e.g., adding leads, updating pipeline, checking open revenue).
  2. Test those commands in a sandbox subaccount and document the exact phrasing that produces reliable results.
  3. Create brand voices so communications stay consistent.
  4. Roll out with guided prompts and a short training session.

Conclusion

This integration is a significant step toward making AI a practical, everyday assistant inside the platform. By combining Ask AI’s conversational capabilities with the MCP server’s safe, auditable access to account actions, we can automate routine tasks, free up time, and maintain consistent communications across teams.

We encourage teams to start small—pick a few routine tasks, test them, and build from there. When done thoughtfully, this tool reduces tech headaches, saves time, and helps teams focus on what matters most: serving customers and growing the business.

"This is absolutely awesome. I can't wait to see how you guys use this."

FAQ

What is the MCP server?

The MCP server is the Model Context Protocol server. It enables the integrated AI to perform account actions by securely interacting with your subaccount’s data and operations. It acts as a controlled bridge that ensures permissions and logs are respected.

Can Ask AI create and edit contacts?

Yes. Ask AI can create new contacts, add notes, and update contact details when provided with the required information (name and phone or email). It uses the MCP server to make those changes directly inside the subaccount.

Does Ask AI charge per action?

Yes, Ask AI consumes one AI credit per successful completed action or response. If the AI cannot complete a requested action, credits are not deducted. Monitor credits so you manage usage costs effectively.

Are irreversible actions protected?

Yes. For any action deemed irreversible or sensitive, Ask AI will request explicit confirmation before proceeding. This safeguard helps prevent accidental data loss or major changes.

Can we customize the assistant’s tone?

Absolutely. Create brand voices inside the platform’s marketing settings. The assistant will use these voices for replies and content generation to maintain consistent messaging.

Will this be available on mobile?

Mobile support is planned and expected to arrive later in the year. When it arrives, the same capabilities will be accessible from mobile devices, enabling field teams to act quickly from anywhere.

What if the AI does something wrong?

Use the thumbs down feedback to flag poor responses. Also, check audit logs and confirm that permissions are configured properly. If needed, limit the scope of allowed actions until confidence grows.

How should we roll this out to clients or staff?

Start with a short training session and a recommended prompts cheat sheet. Limit access where needed and test workflows in a sandbox environment. Gradually expand functionality as confidence and expertise grow.

Where do we find a list of supported actions?

Within the platform’s help documentation, there is a list of supported MCP server actions. Review that list to plan reliable workflows and avoid expecting unsupported actions to work.

How do we ensure privacy and security?

The MCP server follows the platform’s permissioning model and logs actions for auditability. Always limit admin-level actions to trusted users and enable confirmation prompts for sensitive operations.

Read more