Hevo MCP Server Reference

Last updated on Nov 18, 2026

The Hevo MCP server gives your agent the following, which the agent uses to answer your requests:

  • Tools: The actions that the agent can call, such as listing your Pipelines, reading the jobs of a Pipeline, or creating a Destination. Most tools call the Hevo API with your API credentials, and the tools that make changes work only with the scope that you allow.

  • Resources: Read-only information that the agent can read when it needs it, such as the API host of each region and the instructions of the hevo-pipelines skill.

  • MCP prompts: Ready-made tasks that you can start from the menu of your agent, such as a health report of your Pipelines.

The server works without any settings. You can change its behavior with a few environment variables, such as the one that sets the changes that the agent can make.

The following sections list the tools, resources, MCP prompts, and environment variables of the server. To see the endpoints that each tool calls, and whether an endpoint needs your approval, ask your agent to list the Hevo endpoints.


Tools

The server provides three tools of its own, and one group tool for each product, resource, and kind of action. Each group tool calls a group of endpoints of the Hevo API. The agent calls a group tool with the name of the endpoint, the region of your account, and the path parameters, query parameters, and request body of the endpoint. It can read up to 5,000 rows of a list in one call when it reads all the pages of the list. For a larger list, the call returns the rows read so far, and the agent continues from there. A single result is limited to about 1 MB. For a larger result, the agent narrows the request or reads it in pages. With a write, destructive, or delete tool, it can apply the same change to up to 100 resources in one call. If the change needs your approval, you approve all the resources in one prompt.

The server provides the following tools of its own:

Tool Description
list_endpoints Lists the endpoints of the server, with the scope that each endpoint needs and whether it needs your approval.
get_schema Returns the parameters and the request body of an endpoint, and its response when the agent asks for it.
hevo_whoami Checks your connection to Hevo in a region. It reports the region, the name of your key file, the result of a check of your API credentials, the changes that the agent is allowed to make, the folder of your secret files, the version of the server and whether a newer version is available, whether the installed components match the versions with which the release was tested, and the fingerprint of the tools that the server provides. It never shows your API credentials.

The following table lists the group tools for each product and resource, by kind of action:

Product Resource Read Write Destructive Delete
Edge Pipelines, their jobs, and their objects pipelines_read pipelines_write pipelines_destructive pipelines_delete
Edge Destinations destinations_read destinations_write None destinations_delete
Edge Source and Destination types that Hevo supports metadata_read None None None
Standard Pipelines, their objects, schedules, and transformations standard_pipelines_read standard_pipelines_write standard_pipelines_destructive standard_pipelines_delete
Standard Destinations standard_destinations_read standard_destinations_write None None
Standard Models standard_models_read standard_models_write standard_models_destructive None
Standard Workflows standard_workflows_read standard_workflows_write None None
Both Team users and OAuth accounts accounts_read accounts_write None accounts_delete

The destructive tools make the changes that can lose or re-ingest data. They resync an Edge Pipeline or its objects, cancel an Edge job, restart a Standard object, change the position of a Standard Pipeline or object, and reset a Model. They are separate from the write tools, so a setting in your agent that always allows a write tool does not apply to them.

The agent can call the write and destructive tools only with the write or delete scope, and the delete tools only with the delete scope. Before the server asks for your approval, it checks the request body against the schema of the endpoint, so that you are not asked to approve a request with a body that does not match the format of the endpoint. Read Allowing and Approving Changes by your Agent for more information.

The server does not provide the endpoints that return or test sample records of a transformation, as they return records from your Pipelines. As a result, the records that your Pipelines replicate are never sent to your agent.


Resources

A resource is a piece of read-only information that the server makes available to your agent. Unlike a tool, a resource does not call Hevo or change anything. Your agent reads a resource when it needs the information, such as the API host of a region, or the instructions of the hevo-pipelines skill when you do not use the Hevo plugin. In some agents, you can add a resource to the conversation yourself.

The following table lists the resources of the server:

Resource Description
hevo://regions The Hevo regions and their API hosts.
skill://hevo-pipelines/SKILL.md The hevo-pipelines skill, for agents without the Hevo plugin, such as Claude Desktop.
skill://hevo-pipelines/_manifest The list of files in the hevo-pipelines skill.
skill://hevo-pipelines/{path} A reference file of the hevo-pipelines skill, such as references/bulk.md.

MCP Prompts

An MCP prompt is a ready-made task that you can start in your agent without writing the request yourself. The server provides the common tasks of the hevo-pipelines skill as MCP prompts. If your agent supports MCP prompts, you can start these tasks directly. For example, in Claude Code, type / and select the task from the list of commands. The following table lists the MCP prompts:

Prompt Task
hevo_pipeline_health_report Reports the health of every Pipeline over a time window.
hevo_why_is_pipeline_failing Finds the cause of the failure of one Pipeline.
hevo_find_inactive_pipelines Finds the Pipelines that have not run for a period, and the Destinations that no Pipeline uses.
hevo_find_pipelines_for_table Finds the Pipelines that replicate a table or an object.
hevo_event_volume Reports the Event counts of a Pipeline or its objects over time. These counts are not the billed Events.
hevo_summarize_workspace Summarizes the Pipelines and Destinations of a region.

Settings

The server works without any settings. The following table lists the environment variables that the server reads. Except where a row states otherwise, you set them in the same place as the scope, as described in Setting the Scope:

Environment variable Default value Description
HEVO_MCP_DEFAULT_SCOPE read The changes that the agent can make, which can be read, write, or delete. Read Scopes for more information.
HEVO_KEYS_DIR ~/.hevo The folder that contains your key file, hevo-key.json, and the secrets folder. Read Passwords for Sources and Destinations for more information.
HEVO_MCP_LOG_LEVEL INFO The level of detail of the logs, which can be DEBUG, INFO, or WARNING. Set it to DEBUG when Hevo Support asks for a detailed log. The Codex plugin does not pass this variable to the server.

The server does not read your API credentials or your region from environment variables. It reads your API credentials only from your key file, and the agent names your region in every request.


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