MessageGears adds an MCP server, agent-ready APIs and predictive models to its warehouse-native platform
Industry News · Marketing Technology · October 12, 2026
Let the agent build the campaign, and keep the launch button with a person
MessageGears announced on October 8, 2026 a Model Context Protocol (MCP) server, agent-ready APIs and built-in predictive models. Marketers can build audiences, campaigns and content from Claude, ChatGPT or any assistant that supports MCP. The agent cannot launch an email, SMS or push campaign. A person must click launch in the MessageGears interface.
MessageGears is a cross-channel marketing platform. It reads live data in the customer’s cloud data warehouse: Databricks, Snowflake, Google BigQuery or Amazon Redshift. The new tools use the same governed data, so audience counts show current data.
The MCP server comes with 16 built-in skills. The skills teach AI assistants to use MessageGears correctly. Each customer gets a dedicated single-tenant instance. Users sign in with their own credentials and keep their roles and brand permissions. An audit trail records who did each action.
The APIs let agents build audiences, templates and campaigns as drafts. Agents can launch external campaigns on Meta Ads and Google Ads only after an explicit confirmation step. The AI model gets audience definitions and counts. It does not get individual customer records.
MessageGears also offers 10 pre-built predictive models. Jordan Waters, Director of Solutions Engineering, gave an example request. An assistant finds customers with a high churn score and no purchase in 90 days. Then it builds the audience, the campaign and the content.
The takeaway
Give the agent the build work, and keep the send with a person. MessageGears lets an assistant prepare a full campaign. The launch, the customer records and the permissions stay under human control.
Why it matters for agent-led growth
- Separate the build from the launch. Let the agent draft the audience, the content and the launch checklist. Keep a human step for the send, because you cannot recall a sent email.
- Use the permissions that you already have. Each MessageGears user works within existing roles and brand permissions. Give an agent the same role as the person who asks it to work.
- Keep customer records away from the model. The assistant sees counts and definitions only. Ask each vendor which data the model can read.
- Plan for work that starts in the assistant. A marketer can now build a campaign from Claude or ChatGPT. Train the team to start in the assistant and to approve in the platform.
