Hevo MCP Server

Last updated on Nov 18, 2026

The Model Context Protocol (MCP) is an open standard that connects AI agents, such as Claude Code, Claude Desktop, or Codex, to external tools and data sources. A product provides an MCP server, which gives the agent a set of tools that it can call to read or change data in that product.

The Hevo MCP server connects your AI agent to your account. The agent can then view and manage your Pipelines, Destinations, Models, Workflows, team users, and OAuth accounts, based on the requests that you type in the chat of your agent. For example, you can ask your agent, Which of my Pipelines need attention? or Why is Pipeline 12 failing?

With the Hevo MCP server, you can:

  • Check the health of your Pipelines: The agent summarizes your Pipelines, and lists the failing Pipelines and their failed objects.

  • Find the cause of a failure: The agent reads the state, jobs, and objects of a Pipeline, and summarizes what failed, when the failure started, and the next step that you can take.

  • Run routine tasks: The agent pauses, resumes, or runs a Pipeline, or resyncs its objects, when you ask.

  • Create Pipelines and Destinations: The agent asks you for the required details, shows you the request, and creates the Pipeline or the Destination.

  • Review your account: The agent finds the Pipelines that have not run recently, the Destinations that no Pipeline uses, or the Pipelines that replicate a particular table.

Suppose you start your day by checking your Pipelines. You ask your agent, Which of my Pipelines failed in the last 24 hours, and why? The agent reads your Pipelines, their recent jobs, and their failed objects, and returns a table of the Pipelines that need attention, with the failed objects and a suggested next step for each. You then ask the agent to resync one of the failed objects. Before the request reaches Hevo, the server shows you the exact request and its effect, and asks you to approve it. After you approve it, the agent reads the Pipeline again and reports its new state.


Components

Hevo provides two components that work together:

  • MCP server: Provides the tools that the agent calls. The tools that read or change your resources call the Hevo API with your API credentials.

  • Skill: Provides instructions that help the agent use the tools correctly, such as the steps to find the cause of a failure, and the safety rules for changes. The skill is named hevo-pipelines. The Hevo plugin installs the skill in your agent, and the server sends the key rules of the skill to every agent that connects to it.

The server uses your API credentials for every request. As a result:

  • The agent can perform only the actions that your role allows.

  • Each request counts toward the rate limits of your team.

  • Hevo runs the same checks as for any other API request, such as the connection tests for a Source or a Destination.


Safety Controls

The Hevo MCP server protects your Pipelines and your credentials in the following ways:

  • Read-only by default: The agent can view your Pipelines and Destinations, but cannot change them until you allow changes. Read Allowing and Approving Changes by your Agent for more information.

  • Approval for risky changes: Before a delete, a configuration change, a run, or another change that can stop or re-ingest data, the server shows you the exact request in your agent and sends nothing to Hevo until you approve it. The agent cannot approve a request for you.

  • Credentials kept out of the conversation: The server never shows your API credentials to the agent. It reads the passwords for your Sources and Destinations from secret files on your machine when the agent creates or updates them, and replaces passwords and other secrets in every result with a placeholder. Read Hevo MCP Server Security for more information.


Setup Options

The Hevo MCP server runs on your machine, and your requests go from your machine directly to Hevo. The following table lists the agents that you can use with the server, and how to set up the server in each:

Agent Setup Setup Steps
Claude Code The Hevo plugin, which installs the server and the hevo-pipelines skill in one step. Installing the Hevo Plugin in Claude Code
Claude Desktop The hevo-mcp package from PyPI, which you add to the settings of Claude Desktop. Adding the Hevo MCP Server to Claude Desktop
Codex CLI The Hevo plugin, which installs the server and the hevo-pipelines skill in one step. Installing the Hevo Plugin in the Codex CLI
Codex app The Hevo plugin, which installs the server and the hevo-pipelines skill in one step. Installing the Hevo Plugin in the Codex App

Limitations

The Hevo MCP server has the following limitations:

  • The server does not provide your billing or usage information. It can report the number of Events that your Pipelines ingested and loaded, but these counts are not the billed Events. Read Viewing Events Usage for the steps to check your usage in the Hevo UI.

  • The agent cannot create a Source that requires you to sign in through a browser, such as an OAuth sign-in. For a SaaS Source, you must provide a static token or the credentials of your own OAuth app. Read Authenticating SaaS Sources for more information.

  • For Edge Pipelines, a job record reports the status and the counts of the job, but not the reason for a failure. To find the reason for a failure, the agent reads each failed object of the job.

  • The server does not return sample records from your Pipelines, so the agent cannot test transformations.

  • Changes that need your approval work only in agents that can show approval prompts. Claude Desktop, and Claude Code in print mode or in its Visual Studio Code extension, cannot make such changes. Read Approval Prompts in each Agent for more information.

  • The server does not support Activate.


Revision History

Refer to the following table for the list of key updates made to this page:

Date Release Description of Change
Nov-18-2026 2.54 New document.

Tell us what went wrong