Skip to main content
Use MCP when you want an assistant to discover Emailx capabilities and call them with typed inputs. Package 0.11.0 exposes 128 tools through local stdio: 66 reads and 62 writes. The adapter uses the same REST API as your integrations and CLI.

Set up local MCP

Install the adapter in its own Python environment and configure a scoped key.

Explore the tools

Understand each tool’s inputs, permissions and effect.

​
A good first conversation

List my accessible workspaces. Let me select one, then show its LinkedIn campaigns. Do not make any changes.
The assistant should call get_connection, then list_workspaces, then use the selected workspace_id and channel. It should read the relevant campaign before proposing a mutation.

​
What is available

The local adapter does not provide a remote connection URL. Installing it does not make a ChatGPT web connector available. Remote availability explains the separate release boundary. Template edits require templates:write. People import requires contacts:write after a reviewed preview; it does not enroll contacts. Email generation requires campaigns:generate and explicit AI-credit confirmation. Exact-message approval uses messages:approve; email launch requires campaigns:launch-email and a current ready review. Read workflow details in email execution and People import. Host capabilities still gate execution, and a successful activation receipt does not mean delivery.

​
Use MCP or CLI

MCP provides discoverable tools for a conversation. The CLI provides explicit commands, JSON and exit codes for scripts or coding agents. Both preserve workspace, channel, scope and idempotency rules. Workflow skills explain how to complete a job using either interface.