Mc Agent Toolkit

Monte Carlo Agent Toolkit — Claude Code Plugin

Claude Code plugin that brings Monte Carlo data observability into your editor. Bundles skills, hooks, commands, and permissions for all Monte Carlo features in a single plugin.

Requires Python 3.10+.

Installation

Via Marketplace (recommended)

1. Add the marketplace:

   /plugin marketplace add monte-carlo-data/mc-agent-toolkit
   

2. Install the plugin:

   /plugin install mc-agent-toolkit@mc-marketplace
   

3. Updates — claude plugin update pulls in the latest changes.

Manual

git clone https://github.com/monte-carlo-data/mc-agent-toolkit.git
cd mc-agent-toolkit
claude plugin install plugins/claude-code

Restart Claude Code after installing.

Setup

1. Authenticate with Monte Carlo — run /mcp in Claude Code, select the Monte Carlo server, and complete the OAuth flow.
2. Verify — ask Claude: "Test my Monte Carlo connection."

Telemetry

The toolkit sends anonymous skill-usage telemetry by default — which skills are invoked, how often. Each event includes an opaque per-install UUID, a per-session UUID, the skill name, the toolkit version, and the editor it runs in (claude-code). No prompts, arguments, or code are ever sent.

It also sends a Toolkit Installed beacon once per toolkit version — the first time you start Claude Code after installing, and again after each version change — deduped by a local marker, independent of whether you ever run a skill. It carries the install/session UUIDs, toolkit version, and editor (claude-code) — no skill field. Like the skill beacon, it is fail-open and non-blocking.

Authenticated MCP traffic (v1.13.3+). The same anonymous install_id, a per-session id, and the toolkit version also ride as HTTP headers (x-mcd-toolkit-install-id, x-mcd-toolkit-session-id, x-mcd-toolkit-version) on authenticated requests to the Monte Carlo MCP server. This lets the otherwise-anonymous install record be correlated with your account's MCP tool usage server-side — still no prompts, arguments, or code, only the opaque ids and version. The opt-out below disables these headers too.

To opt out, set MC_AGENT_TOOLKIT_TELEMETRY_DISABLED=1 in your shell environment before starting Claude Code. The toolkit will not phone home.

The data is stored in Mixpanel and Datadog and is used only for product-development decisions about which skills to invest in. The UUIDs are generated locally on first session and stored under ~/.claude/mc-agent-toolkit/. Deleting that directory resets your install identity to a fresh anonymous one.

Available Features

Feature	Description	Details
Monitoring Advisor	Analyzes data coverage, creates monitors for warehouse tables and AI agents — covers coverage gaps, use-case analysis, data monitor creation, and agent observability.	Skill README
Prevent	Gates dbt model edits with impact assessments, generates monitors-as-code, and produces targeted validation queries. Full hook enforcement.	Skill README
Generate Validation Notebook	Generates SQL validation notebooks for dbt model changes from a PR or local repo.	Skill README
Push Ingestion	Generates warehouse-specific collection scripts for pushing metadata, lineage, and query logs to Monte Carlo. Includes 10 /mc-* slash commands.	Skill README
Automated Triage	Guides you through automated alert triage — scoring, deep troubleshooting, classification, and actions. Requires extended MCP toolset.	SKILL

The Prevent feature includes PreToolUse hooks that can block edits to dbt SQL files until an impact assessment runs. The hooks only fire on .sql files inside dbt model, macro, or snapshot directories — they do not affect non-dbt files.

To disable the Prevent hooks, set MC_PREVENT_HOOKS_DISABLED=1 in your environment (e.g. in your .zshrc or .bashrc, or via .claude/settings.json under env). The hooks will exit immediately without blocking any edits.

See the Prevent Hook Behavior section in the plugins README for full details on scope and configuration.

Updating

claude plugin update mc-agent-toolkit@mc-marketplace

Uninstalling

claude plugin remove mc-agent-toolkit

The uninstall script restores standalone skill backups if they exist.

Architecture

See the plugins README for the overall plugin architecture, editor support comparison, and hook format details.

Changelog

See CHANGELOG.md.