Generic preview
This is a generic version of this article — vendor-specific item codes
(Datastream, Worldscope, FactSet, and others) are shown as
[VENDOR SPECIFIC ITEM CODE] placeholders.
What MDO Surge Is
Concept Overview
MDO Surge is an AI assistant that knows the mdosnow SDK and your licensed MDO data catalog. You ask a research question in plain English; Surge finds the right data items, verifies they exist, and returns working mdosnow code.
The problem it solves is discovery. MDO exposes hundreds of thousands of
data items across LSEG QAD, S&P Xpressfeed, FactSet, Option Metrics, and
the other integrated vendors, and a general-purpose AI assistant has no
way to know which of them exist, which ones you are licensed for, or
which mdosnow function retrieves them. Left to guess, it invents item
codes such as QAD_PRICE that look plausible and fail at runtime. Surge
does not guess — every item code, index code, and function signature it
returns is looked up first.
Surge runs entirely inside your Snowflake account. It is delivered
as a Snowflake Native App, separate from the MDO SDK app, and is built
on Snowflake Cortex
Agents
and Cortex Search. No data leaves your account boundary, and the agent
holds only the privileges your ACCOUNTADMIN explicitly granted at
install time.
Surge ships as its own Native App, so adopting it is optional — you can run the core MDO SDK without it, and enable Surge later once your security team has reviewed Snowflake’s Cortex features.
What Surge Searches
Surge answers by retrieving before it generates. Cortex Search services index the following, and the agent reads from them on every question:
| Corpus | What it covers |
|---|---|
| MDO documentation | The full published documentation set — API docstrings, vendor tutorials, sample queries, the MDO Style Guide, and the support Q&A library. |
| Custom formulas | Formula names, descriptions, and tags, plus the underlying Python source when you explicitly ask to see an implementation. |
| Index codes | Constituent item codes for building a ConstituentUniverse, filterable by source so index codes and data items stay on the same feed. |
| Powerstrip catalog | Your Powerstrip data dictionary — item codes, descriptions, vendor, source, type, and dimensions. |
| Security identifiers | Vendor-specific security codes, tickers, CUSIP, SEDOL, ISIN, and CIQ trading IDs. |
The Powerstrip and security-identifier searches live inside the MDO SDK app, not the Surge app. That placement is deliberate: it means catalog answers respect your existing licensing, and Surge will not suggest items you are not entitled to.
Surge Tools
When you connect Surge to an AI coding tool over MCP, the following tools become available. Each is served by a Snowflake-managed MCP server, and which app hosts it matters when you write your client configuration.
| Tool | MCP server | Purpose | Hosted in |
|---|---|---|---|
ask_mdo_surge | MDO_SURGE_MCP_SERVER | Documentation, code examples, custom formulas, index codes, and function signatures. | Surge app |
catalog_search | MDO_CATALOG_MCP_SERVER | Find and validate data item codes; confirm vendor, source, type, and frequency. | MDO SDK app |
security_search | MDO_CATALOG_MCP_SERVER | Resolve vendor-specific security identifiers and disambiguate matches by ticker, exchange, or country. | MDO SDK app |
execute_mdo_code, check_mdo_job | MDO_SURGE_CODE_MCP_SERVER | Run generated mdosnow code inside Snowflake and return the resulting rows. Optional; enabled separately by your admin. | Surge app |
Reach for the tools in that order. Use catalog_search and
security_search when you need an exact identifier, and ask_mdo_surge
when you need an explanation, an example, or a formula. Asking Surge for
an item code you could have looked up directly costs more tokens and
takes longer.
Optional: Code Execution
By default Surge writes code and you run it — either locally in a Python environment that has mdosnow installed, or in a Snowflake notebook.
Your administrator can additionally enable Code Execution, which
adds the execute_mdo_code and check_mdo_job tools. With those
enabled, the code Surge generates runs server-side in a Snowpark
Container Service in your own account and returns the resulting rows
directly. The relevant properties:
- There is nothing for individual users to install. No local Python environment is required to get results back.
- Execution runs as the calling user, under that user’s own Snowflake RBAC, and is audited.
- Your data never leaves Snowflake — the code runs next to the data, inside your account boundary.
Code Execution provisions a compute pool and a container service, so it
is opt-in rather than part of the base install. The tools also live on a
separate MCP server (MDO_SURGE_CODE_MCP_SERVER), which means a
user only gets execution if they add that connector explicitly.
Getting Access
Surge requires a one-time setup by a Snowflake ACCOUNTADMIN before
anyone can use it. That work installs the Surge Native App, grants it
Cortex access, and builds the Cortex Search services and MCP servers.
Once setup is complete, there are two ways to use Surge:
- Using Surge in Snowflake — ask questions in Snowflake Intelligence in Snowsight. Nothing to install.
- Connecting Surge over MCP — wire Surge into Claude Code, Codex, or GitHub Copilot so your coding assistant can look up items and generate code as you work.
If Surge is not yet available in your account, contact MDO support at support@mydataoutlet.com and ask for the MDO Surge Admin Installation Guide.