CampaignHQ connects an AI assistant to your marketing account through MCP, the Model Context Protocol. It lets permitted users inspect reports, preview audiences and prepare email, WhatsApp and SMS work. Campaign launch and automation activation remain human actions. Access is currently in beta, with gradual enablement for individual eligible company accounts.

Last reviewed: October 10, 2026.

If your team already uses AI to write emails, the next question is practical: can it work with your campaign data without giving it permission to launch a campaign? CampaignHQ’s MCP server separates those jobs. An assistant can help prepare the work; a person reviews the campaign in CampaignHQ before it goes out.

This guide explains the documented capabilities and controls for marketing and retention teams. It is not an installation manual or a promise that every account already has access. Use the current CampaignHQ MCP documentation for setup and the complete tool reference.

What MCP changes for a marketing team

The Model Context Protocol is a standard way for AI applications to connect to data and tools. In this setup, your AI assistant is the client and CampaignHQ provides the server. The assistant requests a documented operation, CampaignHQ applies your permissions, and the result returns to the assistant.

That is different from pasting a campaign screenshot into a chatbot. A connected assistant can retrieve the permitted campaign report or inspect the fields available for segmentation. It is also different from a new predictive marketing model: connecting an assistant does not establish that it can forecast purchases, improve deliverability or choose a winning offer.

The commercial reason to evaluate it is less manual movement between your data and your planning work. For a team coordinating email and WhatsApp customer journeys, that means asking better questions about the same audience, campaign and follow-up process, rather than producing more disconnected copy.

Which AI apps can connect?

CampaignHQ documents connections for Claude, ChatGPT, Codex and other compatible clients. Some use a browser sign-in; others use a scoped API key. Your client plan, administrator settings and supported authentication method can affect setup. A documented configuration is not a guarantee that every version of every client has been tested.

First confirm that MCP is enabled for your company. Then use the regional server shown under CampaignHQ’s Connect AI settings. Do not copy a colleague’s key or assume their access grants you the same role. Keep keys out of shared prompts, screenshots and repositories.

If ChatGPT is your team’s starting point, the ChatGPT email marketing guide covers the email-specific workflow. This page focuses on what CampaignHQ exposes, regardless of the assistant you choose.

What the assistant can read and analyse

Account and campaign performance

The account-context tool identifies the connected company, role, permissions, time zone and channels. It should be the first check. Account and campaign reporting then provide recorded performance for the requested period.

Campaign reports distinguish email activity, WhatsApp delivery stages and SMS counts and rates. The documented email report includes unique and bot activity; WhatsApp reporting includes delivery events and recorded Meta error codes. Use those distinctions when interpreting results. A reported click, read or attributed sale is not automatically an incremental purchase caused by the message.

Audience rules and contact context

The assistant can inspect available lists, segments, fields, tags and tracked events. It can preview proposed segment rules without saving them, returning a count and masked examples. That is useful when “recent customer” or “inactive subscriber” has become an ambiguous instruction.

A preview is not consent. Your team still needs a valid basis for contacting the audience and the correct channel eligibility. WhatsApp’s Business Messaging Policy continues to apply when AI helps prepare the campaign.

Delivery and journey investigation

Delivery diagnostics expose recorded contact reachability, suppression information, recent message events and WhatsApp errors. Automation reports show how contacts move through steps. Chat-flow reports show executions and node-level results; the documented daily rollups exclude today.

These tools help you locate a question worth investigating. They do not prove the root cause of every failure or repair a blocked number. A report can identify where evidence stops; it should not fill that gap with a confident story.

Product channels and MCP coverage are different

CampaignHQ supports Email, WhatsApp, SMS and RCS (Rich Communication Services) as product channels. Its current MCP catalogue documents email, WhatsApp and SMS operations. RCS product support does not establish RCS tools in the connector: confirm the available RCS workflow in a demo rather than asking the assistant to draft, report on or send an RCS campaign.

SMS reports, campaign drafts and templates

For SMS, begin with account context to identify the connected service. Search for the campaign, then retrieve its report for recorded counts and rates. Check which metrics are returned; do not import email open-rate assumptions into an SMS review.

To prepare an SMS campaign, select the intended audience, service_id and template_id. Template tools can retrieve the body and DLT identifier, or create and edit templates. Where India’s Distributed Ledger Technology (DLT) registration applies, supply the registered template identifier and verify its review status. An edit can send a template back for review.

The CampaignHQ SMS product page describes campaigns, customer-event triggers and transactional alerts. Those are platform capabilities, not permission for an AI assistant to launch them. SMS tests run in the app, not through the self-test email tool; a person reviews and launches the campaign.

What the assistant can create or change

With the applicable write permission, CampaignHQ documents campaign drafts, templates, saved segments and contact updates. These operations are not all harmless drafts. Changing a segment or a contact changes account data, so inspect the proposed change before approving it in your AI client.

• Campaign drafts: prepare email content, a WhatsApp campaign using an approved template and its variables, or an SMS campaign using a selected service and template. Saving a schedule on a draft does not launch it.
• Campaign edits: update supported draft fields. A visual-editor email keeps its design; content editing may need to happen in the app.
• Segments: preview first, then save. Updating an existing segment replaces its rules rather than appending an extra condition.
• Automations: create a new draft for review. The current documentation sends editing of an already-saved automation back to the app.
• WhatsApp chat flows: create a draft or change a draft/inactive flow. Review the graph, number and handoff before activating it.

A chat flow is not automatically an autonomous AI support agent. The documented builder creates a flow with defined steps and branches. Evaluate the actual customer experience, not the label.

Where human control remains

CampaignHQ’s launch-readiness check reports blockers, audience information and a review link. It does not launch or schedule the campaign. The activation-readiness check for an automation or chat flow also reports problems without activating anything. A person completes those actions in CampaignHQ.

There are two important messaging exceptions. A connected assistant can send an email test only to the signed-in user’s own email address. It can also send an individual WhatsApp inbox reply when the required send scope is granted and an administrator has enabled that capability. Therefore, “the MCP can never send any message” would be inaccurate.

The inbox-reply path has its own assignment, subscription and messaging-window checks. Do not give a reporting assistant reply permission merely because it is available. Keep customer-facing replies outside an initial reporting pilot unless the team has explicitly approved that use.

The documented pause tool can stop some sending operations, but it is not an undo button for messages already sent. In particular, the documentation says an already-sending WhatsApp or SMS campaign cannot be paused through that tool. Review launch decisions before, not after, reaching customers.

Permissions, privacy and offboarding

The connection acts as the user, within their role. Read access, general write access, inbox reading, inbox management and inbox sending are separate controls. An inbox manager does not automatically need permission to send through AI.

Administrators can disable AI-agent access company-wide. Users can disconnect connected apps or revoke API keys. Build that into offboarding: removing a colleague from an AI workspace is not a substitute for checking the marketing-platform connection too.

Regional CampaignHQ endpoints do not prove that every external AI provider processes data in that same region. The assistant receives results under the connection’s permissions, and its provider’s terms govern its subsequent handling. Review those terms, data-retention settings and your internal approval requirements before exposing customer data.

This is a broader operating issue, not a CampaignHQ-only concern. Snowflake’s marketing governance discussion describes the tension between useful access and accountable action. For your team, make the rule concrete: name the person responsible for each write and each launch.

How to evaluate access without a platform migration

1. Choose one existing job. Use a campaign review or an audience preview, not “automate all marketing.”
2. Confirm availability and access. Check company enablement, the connected identity and the minimum permissions for that job.
3. Compare the answer with the app. Check the same date range, campaign, filters and metric definitions.
4. Inspect a draft before trusting a workflow. Verify audience, sender, content, variables and the review link.
5. Measure operator value. Record preparation effort, corrections and review quality. Treat campaign outcomes as a separate test.

If an existing platform’s connector already solves that job, a migration may be unnecessary. Compare AI email marketing tools by the work they support, not by whether they display an AI badge.

CampaignHQ identifies itself as a Meta Tech Provider and supports Email, WhatsApp, SMS and RCS customer engagement. The reason to request a demo is to test that operating fit with your audience and team, not to assume MCP alone will improve retention.

Frequently asked questions

Does MCP replace CampaignHQ?

No. It provides another interface to documented CampaignHQ operations. Your account, permissions, campaign review and launch process remain in CampaignHQ.

Can every CampaignHQ customer connect immediately?

No. The current FAQ describes beta access and gradual account enablement. Confirm your company’s availability before planning a rollout.

Does connecting AI use CampaignHQ AI credits?

The MCP FAQ says usage from an external AI app does not draw on the CampaignHQ AI allowance. Your external AI app’s own plan and charges are separate.

Can an assistant resubscribe an unsubscribed contact?

The documented unsubscribe operation cannot be reversed through that tool. Resubscription needs the contact or a person in the app; do not treat an AI instruction as renewed consent.

What should we bring to an MCP demo?

Bring a real reporting or campaign-preparation problem, your channel mix, required permissions and the person who approves launches. Use sanitised examples until your data-access policy is settled.

Written by CampaignHQ Team