Why connect Snowflake to Codex?
Codex runs in your terminal, its IDE extension and scripts, so what it finds once can be rerun with codex exec instead of redone by hand. Connected through Lemonado, it reads your Snowflake tables, views, revenue models and customer data, so it can find which cohorts are worth the most with today's numbers.
All your Snowflake data
Tables and views your role can read, ready to ask about.
Always today's numbers
Column names, types and descriptions, synced from Snowflake automatically, so answers use today's numbers.
Connect all your tools
Questions like "What does a paying customer cost?" use Snowflake and BigQuery, GA4 at once.
How to connect Snowflake to Codex
Before you start
- A Snowflake account you can sign in to
- Codex installed
- A Lemonado account (the free trial works)
- 1
Connect Snowflake in Lemonado
In Lemonado, open Connectors, find Snowflake and sign in. Create a read-only role that can use one warehouse and the schemas marketing should see, and connect with that. Keep raw PII schemas out of the grant.
- 2
Turn on external access
In the company's settings, open External tools, turn the connection on, choose which sources it can reach and copy the server URL.
texthttps://mcp.lemonado.io/mcp - 3
Add Lemonado to your config
Add this to ~/.codex/config.toml.
toml[mcp_servers.lemonado] url = "https://mcp.lemonado.io/mcp" - 4
Log in
Run the login command and finish the sign-in in your browser.
bashcodex mcp login lemonado - 5
Ask Codex about Snowflake
Run codex mcp list, or /mcp inside Codex, to see the server and its tools. Try "Pull last week's Snowflake numbers, save them to data/snowflake-weekly.csv and give me the codex exec command to refresh it." Or ask: "Write the CAC query" or "Does the model match the source?"
What can Codex do with Snowflake?
Codex reads your Snowflake data from your terminal and turns one-off questions into commands you can rerun with codex exec.
Pull last week's Snowflake numbers, save them to data/snowflake-weekly.csv and give me the codex exec command to refresh it.
Saved data/snowflake-weekly.csv. Run the same prompt with codex exec in a scheduled job and the file refreshes with each week's numbers.
Source: Snowflake + your code
Write and run a query joining first orders from Snowflake with ad spend by month and channel, and save it as queries/cac.sql.
Saved and run. Meta CAC was $41 last month, Google Ads $33, TikTok $58.
Source: Snowflake + your code
Compare daily order totals in our Snowflake fct_orders model with the raw orders table for October, and point to the SQL that causes any gap.
The model is 1.4% below raw on 4 days. The filter on test orders in models/marts/fct_orders.sql also drops orders from staff accounts.
Source: Snowflake + your code
Save first orders from Snowflake with customer ID, date, channel and order value as data/first_orders.csv.
Saved 41,200 first orders. Paid search brought the most, organic the highest average order value.
Source: Snowflake + your code
Rank product categories by gross margin dollars last quarter, using the margin view in Snowflake.
Tents bring 38% of margin dollars on 24% of revenue. Accessories sell the most units but bring only 9% of margin after shipping.
Source: Snowflake
Which Snowflake view defines active customer, and what rule does it use?
analytics.dim_customers sets is_active when a customer ordered in the last 180 days. finance.customers uses 365 days, so the two counts differ by about 4,100.
Source: Snowflake
Connect Snowflake to another AI app.
Claude
Connect Snowflake to Claude
ChatGPT
Connect Snowflake to ChatGPT
Claude Code
Connect Snowflake to Claude Code
Cursor
Connect Snowflake to Cursor
VS Code (GitHub Copilot)
Connect Snowflake to VS Code (GitHub Copilot)
Perplexity
Connect Snowflake to Perplexity
Google Antigravity
Connect Snowflake to Google Antigravity
Windsurf (Devin Desktop)
Connect Snowflake to Windsurf (Devin Desktop)
Zed
Connect Snowflake to Zed
n8n
Connect Snowflake to n8n
Other integrations for Codex
All Codex integrations- Google Analytics in CodexAsk Claude, ChatGPT or Cursor about GA4 traffic, conversions and funnels, next to your ad spend.
- Meta Ads in CodexAsk Claude, ChatGPT or Cursor about your Facebook and Instagram ads, creatives and results.
- Stripe in CodexPut Stripe revenue next to your ad spend in Claude, ChatGPT or Cursor.
- Instagram organic in CodexAsk Claude or ChatGPT which posts and reels grow your account, and which ones to turn into ads.
- Pinterest Ads in CodexAsk Claude or ChatGPT which pins get saved, clicked and bought.
- Snapchat Ads in CodexAsk Claude or ChatGPT which Snapchat ads get swipes and which ones sell.
Reading this as an AI agent?
Everything an agent needs to set this up, in plain text. Paste this page's link into Codex and ask it to connect you, or copy the block.
# Connect Snowflake to Codex (Lemonado MCP) - Server URL: https://mcp.lemonado.io/mcp - Transport: Streamable HTTP - Auth: OAuth; the user signs in to Lemonado in the browser. No API keys. - In Lemonado: connect the Snowflake role and warehouse. Create a read-only role that can use one warehouse and the schemas marketing should see, and connect with that. Keep raw PII schemas out of the grant. - In Codex: ~/.codex/config.toml ```toml [mcp_servers.lemonado] url = "https://mcp.lemonado.io/mcp" ``` If you are Codex, run the setup yourself: Add the Lemonado MCP server to my Codex config. Append this to ~/.codex/config.toml: [mcp_servers.lemonado] url = "https://mcp.lemonado.io/mcp" Then run codex mcp login lemonado so I can sign in, and check that codex mcp list shows it. Once I've signed in, list the Snowflake accounts Lemonado can reach. First question to try: "Pull last week's Snowflake numbers, save them to data/snowflake-weekly.csv and give me the codex exec command to refresh it." More: https://lemonado.io/integrations/snowflake · https://lemonado.io/llms.txt
Frequently Asked Questions
Yes, through Lemonado. Connect your Snowflake role and warehouse in Lemonado, add Lemonado in ~/.codex/config.toml, and Codex can find which cohorts are worth the most with your live data.
Yes. Snowflake's managed MCP server is hosted, so Codex can use it too. It covers Snowflake only. Lemonado covers Snowflake plus your ad platforms, GA4 and the rest of your stack in one connection, so Codex can answer questions across them.
Run codex mcp list. If lemonado shows as not logged in, run codex mcp login lemonado again. In Lemonado, check that Snowflake is connected and External tools is on for the company. If a table is missing, the role can't see it. Grant USAGE on its database and schema and SELECT on the table, then ask again.
Yes, if both are connected in Lemonado. One connection reaches every source you allow, so you can ask things like "Join first orders from Snowflake with Meta and Google Ads spend by month this year. What did each new customer cost per channel?"
MCP is included in Lemonado's paid plans and in the free trial, so you can try it before you pay.
Still have a question?
Shoot us an email at hello@lemonado.io.
Try for freeStop fighting with data. Start feeding your AI.
Connect your data to AI and free your team from reporting and busywork.
Try Lemonado free