Quickstart

Four steps. Everything runs inside your own agent, on your own machine.

Nothing reaches Nodal. The open-source package is instructions and local scripts that run inside your agent — no data, no credentials, no queries, no results, no telemetry. Warehouse access goes through an MCP server you configure with a read-only identity. The only external data flow is the one you already have with your agent’s model provider. The hosted MCP is a separate, paid product that activates only when you opt in.

Install the plugin

Pick one installation method per host — installing both a native plugin and skills.sh copies makes every skill appear twice.

Claude Code
Codex
Cursor & other skill-compatible agents

claude plugin marketplace add nodal-data/nodal-context
claude plugin install nodal-analytics@nodal

To update an existing installation:

claude plugin marketplace update nodal
claude plugin update --scope user nodal-analytics@nodal

codex plugin marketplace add nodal-data/nodal-context
codex plugin add nodal-analytics@nodal

npx skills@latest add nodal-data/nodal-context

This copies editable skills into your project. Select setup-nodal along with the workflows you want to use.

Or let your agent install it for you

If you already use an AI coding agent, paste this into a project where you want to evaluate Nodal. The agent guide it reads teaches the agent to pick one installation path, ask before changing anything, and stop at the new-session boundary.

Help me understand and set up Nodal Analytics for this project. Read
https://github.com/nodal-data/nodal-context/blob/main/docs/agent-guide.md
first. Explain what will change, choose only one supported installation method,
and ask before installing anything or editing configuration. Then walk me
through it one step at a time.

Connect a read-only warehouse over MCP

Wire your warehouse’s MCP server into your agent with a read-only role — Nodal only ever SELECTs. The Connect your database page lists the vendor-maintained MCP servers for Snowflake, BigQuery, Redshift, and Databricks, plus the optional query-history grant that makes the interview’s best input available.Start your agent from the folder that holds your data lineage, so the context repo Nodal creates sits next to it:

code/                        <-- start Claude Code / Codex here
├── dbt-repo/                # your existing dbt project (optional, recommended)
└── analytics-context/       # created by Nodal, a sister repo to the dbt project

Start a new session and run setup

Plugins are discovered at the session boundary, so start a fresh agent session after installing — a skill that “isn’t showing up” is almost always this. Then run setup once:

# Claude Code
/nodal-analytics:setup-nodal

# Codex
$setup-nodal

Setup probes your warehouse connection (read-query, metadata, query-history), discovers nearby dbt and context sources, and writes only sanitized paths and capability classifications to a gitignored .nodal.local.json — never credentials.

Ask

Either take the short path or the full one:

# ~30 minutes — one domain, five high-leverage questions, a live check:
Take Nodal for a test drive on one analytics domain.

# The complete interview:
Build my analytics context.

Both write a reviewable ../analytics-context/ git repo and offer to push it to your own private remote. Every confirmed definition also becomes an eval seed, and each domain closes by verifying its answers against a dashboard you trust.

Build your context

What the interview asks, what it needs, and what it writes.

Evaluation as you build it

Watch a live dashboard verification and see the accuracy delta.

Share with your team

Put the context in front of everyone’s agent with one .mcp.json file.

Get started

Open source (free)

Nodal MCP (paid)

Enterprise

Help

Next

Build your context

Evaluation as you build it

Share with your team

​Next

Build your context

Evaluation as you build it

Share with your team

Next