Artie MCP: Manage Pipelines with Agents

Artie MCP is a remote MCP server with 48 reviewed tools for working with CDC pipelines from your AI client. Create or operate pipelines, select tables, start or pause replication, run approved backfills, inspect schema changes, manage private connectivity, and search the data catalog.

Artie MCP works with Claude Code, Cursor, Codex, and other OAuth-capable MCP clients.

Install Artie MCP

Claude Code and Cursor users can install the Artie plugin, which includes the pipeline-setup skill for guided pipeline creation. Codex and other compatible clients can connect directly to the Artie MCP server using its URL and OAuth.

On first use, sign in, link an environment, and approve the requested scopes. You do not need an API key.

Claude Code

Install the plugin to register the server and the Artie skill in one step.

claude plugin marketplace add artie-labs/artie-mcp
claude plugin install artie@artie-mcp

Or add the server without the plugin:

claude mcp add-json artie '{"type":"http","url":"https://mcp.artie.com/mcp"}'

Cursor

Install the Artie plugin from the Cursor Marketplace.

Or add the server directly to .cursor/mcp.json:

{
  "mcpServers": {
    "artie": {
      "url": "https://mcp.artie.com/mcp"
    }
  }
}

Codex

Codex CLI, the Codex IDE extension, and the ChatGPT desktop app share one MCP config. Add the server to ~/.codex/config.toml:

[mcp_servers.artie]
enabled = true
url = "https://mcp.artie.com/mcp"

Restart Codex, then run /mcp to confirm that artie is listed and authenticate when prompted.

Other clients

Point any OAuth-capable MCP client to:

https://mcp.artie.com/mcp

Read the Current State Before You Change It

Most pipeline work starts with what exists now, not with a configuration file from last quarter.

Try these prompts:

List the production pipelines in this environment.

Show the schemas and tables on our saved Postgres connector.

Which tables on the orders pipeline are still backfilling?

The server can return pipeline details, saved connector configuration, databases, schemas, tables, table metadata, table-level usage statistics, and full-text data-catalog search. Sensitive credential fields remain masked.

Operate Pipelines From Your Client

Use Artie MCP to move from a live pipeline question to an approved change:

Create a pipeline from the staging Postgres connector to Snowflake.

Detect schema changes on the staging pipeline.

Some operations begin background work. A schema-change check is queued before results are available, and a backfill runs against tables already attached to a running pipeline. Inspect the returned status before treating that work as complete.

Artie MCP organizes its 48 tools across 10 resource groups: pipelines, connectors, source readers, SSH tunnels, PrivateLink connections, column hashing salts, encryption keys, ingestion API keys, the data catalog, and company-wide schema reconciliation. See the MCP tools reference for the complete list.

The Skill Handles Sequencing

MCP provides authorized access. It does not tell an agent what order to use those permissions in, and pipeline setup is order-dependent.

The pipeline-setup skill handles that sequence. It validates and pings connectors before creating them, preserves full pipeline configuration during updates, asks for confirmation before the write that saves the plan, and stops on a failed gate rather than guessing the next action.

Try Artie MCP

Install the plugin, or configure your MCP client with https://mcp.artie.com/mcp.