Skip to content
→Connect Snowflake — free
GoMarble MCP

MCP for Snowflake

GoMarble MCP connects AI assistants to Snowflake through a query interface, letting them run SQL against your warehouse from a natural-language request — and join that answer to ad platform data on the same connection.

One tool, and an unusually broad one: anything expressible as a query against data you can already reach.

Pranshu in ↗

Growth, GoMarble · Updated 2026-10-09 · 7 min read

Get Started with GoMarble MCP View MCP Documentation →

What is MCP for Snowflake?

M

C

P

f

o

r

S

n

o

w

f

l

a

k

e

m

e

a

n

s

c

o

n

n

e

c

t

i

n

g

a

n

A

I

c

l

i

e

n

t

t

o

y

o

u

r

S

n

o

w

f

l

a

k

e

w

a

r

e

h

o

u

s

e

t

h

r

o

u

g

h

t

h

e

M

o

d

e

l

C

o

n

t

e

x

t

P

r

o

t

o

c

o

l

,

s

o

t

h

e

a

s

s

i

s

t

a

n

t

c

a

n

r

u

n

q

u

e

r

i

e

s

o

n

y

o

u

r

b

e

h

a

l

f

.

U

n

l

i

k

e

t

h

e

p

l

a

t

f

o

r

m

c

o

n

n

e

c

t

i

o

n

s

,

w

h

i

c

h

e

x

p

o

s

e

f

i

x

e

d

e

n

d

p

o

i

n

t

s

,

t

h

i

s

i

s

a

q

u

e

r

y

i

n

t

e

r

f

a

c

e

—

i

t

s

s

c

o

p

e

i

s

w

h

a

t

e

v

e

r

y

o

u

r

w

a

r

e

h

o

u

s

e

c

o

n

t

a

i

n

s

a

n

d

y

o

u

r

c

r

e

d

e

n

t

i

a

l

s

p

e

r

m

i

t

.

How to connect Snowflake to GoMarble MCP

Connect Snowflake once in GoMarble, then point any supported AI client at the MCP server.

  1. S
  2. i
  3. g
  4. n
  6. u
  7. p
  9. o
  10. n
  12. G
  13. o
  14. M
  15. a
  16. r
  17. b
  18. l
  19. e
  20. ,
  22. c
  23. o
  24. n
  25. n
  26. e
  27. c
  28. t
  30. S
  31. n
  32. o
  33. w
  34. f
  35. l
  36. a
  37. k
  38. e
  40. u
  41. n
  42. d
  43. e
  44. r
  46. I
  47. n
  48. t
  49. e
  50. g
  51. r
  52. a
  53. t
  54. i
  55. o
  56. n
  57. s
  58. ,
  60. t
  61. h
  62. e
  63. n
  65. a
  66. d
  67. d
  69. t
  70. h
  71. e
  73. G
  74. o
  75. M
  76. a
  77. r
  78. b
  79. l
  80. e
  82. M
  83. C
  84. P
  86. s
  87. e
  88. r
  89. v
  90. e
  91. r
  93. a
  94. s
  96. a
  98. c
  99. u
  100. s
  101. t
  102. o
  103. m
  105. c
  106. o
  107. n
  108. n
  109. e
  110. c
  111. t
  112. o
  113. r
  115. i
  116. n
  118. y
  119. o
  120. u
  121. r
  123. A
  124. I
  126. c
  127. l
  128. i
  129. e
  130. n
  131. t
  132. .
  134. T
  135. h
  136. e
  138. s
  139. a
  140. m
  141. e
  143. c
  144. o
  145. n
  146. n
  147. e
  148. c
  149. t
  150. i
  151. o
  152. n
  154. s
  155. e
  156. r
  157. v
  158. e
  159. s
  161. e
  162. v
  163. e
  164. r
  165. y
  167. c
  168. l
  169. i
  170. e
  171. n
  172. t
  174. y
  175. o
  176. u
  178. a
  179. d
  180. d
  181. .

Connecting to Claude

The current steps, from the GoMarble MCP page:

  1. Open Settings, then Connectors.
  2. Choose Add custom connector.
  3. Enter https://apps.gomarble.ai/mcp-api/sse, then select Authorize.
  4. Choose the GoMarble workspace and accounts Claude can use.

Authorization uses GoMarble’s consent and workspace permissions, completed in the browser window your client opens.

Other supported clients

GoMarble MCP is a remote MCP server at https://apps.gomarble.ai/mcp-api/sse, currently supported in Claude, ChatGPT, Cursor, n8n, Hermes and OpenClaw, plus other clients that support compatible remote or local MCP servers. Setup differs slightly per client, so follow the canonical instructions rather than adapting the Claude steps.

View the complete GoMarble MCP setup guide →

What the free plan includes

  • Up to $50,000 in cumulative ad spend
  • One account per data source
  • No edit actions — read and analysis only

What can you do with GoMarble MCP for Snowflake?

Grouped by the job rather than by the tool that does it.

Analyze

  • Run queries against your warehouse from a plain-language request
  • Pull business data that never reaches the ad platforms — margin, LTV, returns, offline revenue
  • Aggregate across tables the ad platforms cannot see

Investigate

  • Check what the warehouse says against what a platform reports
  • Reconcile platform-attributed revenue with booked revenue
  • Interrogate the data model behind a number you do not trust

Act

  • Scope depends on the credentials you connect with — treat it as you would any warehouse access

GoMarble MCP capabilities for Snowflake

CapabilityWhat GoMarble MCP can do
Query executionRuns SQL against Snowflake from a natural-language request
ScopeWhatever your warehouse holds and your connected credentials permit — not a fixed endpoint list
Business contextMargin, lifetime value, returns and offline revenue that ad platforms never see
ReconciliationCompare platform-reported figures against the numbers the business actually books
Access controlGoverned by the Snowflake role you connect with, not by this connection

What data can GoMarble MCP access?

The data available for Snowflake, grouped by type.

Warehouse

Any table or view your connected role can read

Typical use

Order and revenue tables · Customer and LTV models · Margin and cost data · Offline conversions · Marketing data already piped in

GoMarble MCP workflows for Snowflake

Jobs marketers actually want done, rather than features. Each one is a single natural-language request.

Judge campaigns on margin, not revenue

“Which Meta campaigns drive the most gross margin, not just revenue?”

What GoMarble MCP does: Pulls campaign revenue from Meta and joins it to margin data in the warehouse — a question neither source can answer alone.

What you get:

Reconcile platform and booked revenue

“How does Meta-attributed revenue compare to what we actually booked last month?”

What GoMarble MCP does: Queries booked revenue in Snowflake and compares it against platform-reported figures.

What you get:

Bring LTV into the CPA conversation

“What is our allowable CPA by segment based on actual LTV?”

What GoMarble MCP does: Queries the LTV model and sets it against current CPA by channel.

What you get:

Account for returns

“Which channels look worse once returns are deducted?”

What GoMarble MCP does: Pulls return rates by channel from the warehouse and recalculates efficiency.

What you get:

Check a number you do not trust

“Where does this revenue figure actually come from?”

What GoMarble MCP does: Interrogates the underlying tables rather than taking a dashboard at face value.

What you get:

Pull offline conversions

“How many of last month's leads actually closed?”

What GoMarble MCP does: Queries the closed-won data that never makes it back to the ad platforms.

What you get:

Example GoMarble MCP prompts for Snowflake

Things real marketers ask, grouped by intent. Paste any of them once connected.

Revenue and margin

  • “What was booked revenue by channel last month?”
  • “Which products carry the best margin?”
  • “Show me revenue net of returns by channel”
  • “How does gross margin split across paid and organic?”

Customer

  • “What is average LTV by acquisition channel?”
  • “How does repeat rate differ by first-order source?”
  • “Which segments have the highest lifetime value?”

Reconciliation

  • “Compare Meta-attributed revenue with booked revenue”
  • “Does platform ROAS match what the warehouse says?”
  • “Where is the gap between reported and actual conversions?”

Offline

  • “How many leads from last month closed?”
  • “What is the lead-to-customer rate by channel?”
  • “Show me offline revenue by source”
  • “Which channels produce leads that actually close?”

Cohorts and trend

  • “How has repeat purchase rate moved over the last four quarters?”
  • “Show me revenue by acquisition cohort”
  • “Which months produced the most valuable customers?”

Why use GoMarble MCP for Snowflake?

An API gives AI access to data. GoMarble MCP gives it the marketing context and workflows needed to work with that data.

GoMarble MCP for cross-platform marketing analysis

Marketers rarely decide anything from one platform. GoMarble MCP can give AI a view of the wider marketing system rather than forcing it to reason from one disconnected source at a time.

CombinationWhat becomes answerable

How does GoMarble MCP work?

StepWhat happens
1You ask a question in natural language
2The AI client works out what information it needs
3GoMarble MCP retrieves the relevant marketing data and tools
4GoMarble supplies the marketing context around it
5The AI analyses the result
6You get an answer, a recommendation, a report — or a proposed action

Write access is plan-dependent. On the read-only plan you can connect a client and analyse supported accounts without making changes. Where write access is enabled, changes are proposed rather than applied directly — every write tool in the server is a propose call, and consequential actions route through the permissions and approval gates configured for your workspace.

Who is GoMarble MCP for?

Analytics teams

Answer an ad-hoc warehouse question without opening a SQL client.

Performance leads

Bring margin and LTV into channel decisions rather than optimising to platform revenue.

Finance-adjacent marketers

Reconcile what platforms claim against what the business booked.

GoMarble MCP vs traditional Snowflake workflows

Traditional workflowWith GoMarble MCP

Frequently asked questions

What is MCP for Snowflake?
It means connecting an AI client such as Claude or ChatGPT to your Snowflake warehouse through the Model Context Protocol, so the assistant can run queries for you. Unlike the platform connections, it is a query interface rather than a fixed set of endpoints.
Can it write to my warehouse?
Its scope is governed by the Snowflake role you connect with, not by this integration. Connect a read-only role if you want it strictly read-only — that is the correct control point, and the same discipline you would apply to any other warehouse client.
Do I need to know SQL?
Not to ask a question, but it helps to know what is in your warehouse. The assistant writes the query; you still need enough context to judge whether the answer is reasonable and whether it queried the table you meant.
Why connect a warehouse to an ad tool at all?
Because ad platforms optimise to what they can see, and what they cannot see is margin, returns and lifetime value. Joining those two sides is how a ROAS conversation becomes a profit conversation, and it is hard to do when the data lives in separate tools.
Is it safe to let an assistant query the warehouse?
Treat it exactly as you would any other client with those credentials. The access is whatever the connected role permits, so scope the role deliberately rather than relying on the tool to restrain itself.
Which AI clients work with it?
Claude, ChatGPT, Cursor, n8n, Hermes and OpenClaw. The same server URL works for all of them.
What does the free plan allow?
Read and analysis only, one account per data source, and a cap of $50,000 in cumulative ad spend.
Does it replace our BI tool?
No. BI is for the numbers many people check repeatedly, and it should stay. This is for the one-off question, and particularly the kind that joins warehouse data to ad data — which is usually awkward to model in advance.

Why GoMarble MCP?

The work you stop doing:

  • T
  • h
  • e
  • w
  • a
  • r
  • e
  • h
  • o
  • u
  • s
  • e
  • a
  • n
  • d
  • t
  • h
  • e
  • a
  • d
  • p
  • l
  • a
  • t
  • f
  • o
  • r
  • m
  • s
  • a
  • n
  • s
  • w
  • e
  • r
  • a
  • b
  • l
  • e
  • i
  • n
  • o
  • n
  • e
  • c
  • o
  • n
  • v
  • e
  • r
  • s
  • a
  • t
  • i
  • o
  • n
  • ,
  • s
  • o
  • m
  • a
  • r
  • g
  • i
  • n
  • a
  • n
  • d
  • l
  • i
  • f
  • e
  • t
  • i
  • m
  • e
  • v
  • a
  • l
  • u
  • e
  • c
  • a
  • n
  • i
  • n
  • f
  • o
  • r
  • m
  • a
  • m
  • e
  • d
  • i
  • a
  • d
  • e
  • c
  • i
  • s
  • i
  • o
  • n
  • i
  • n
  • s
  • t
  • e
  • a
  • d
  • o
  • f
  • a
  • r
  • r
  • i
  • v
  • i
  • n
  • g
  • a
  • w
  • e
  • e
  • k
  • l
  • a
  • t
  • e
  • r
  • .

Connect your marketing stack once. Give AI the context it needs. Ask in plain English, and let it analyse, explain, recommend and — where enabled — propose the change.

Your Snowflake data is already there. Now give AI access to it.

Connect GoMarble MCP to your marketing stack and start analysing Snowflake from the AI tool you already use.